FormalPara Key Points

Different definitions are used to define psoriasis relapse after treatment discontinuation, which limits the comparison between studies.

Biological agents are associated with a longer time to relapse than oral systemic agents.

Among biologicals, interleukin-23 antagonists are associated with the longest time to relapse.

1 Introduction

Biologics for psoriasis generally promote excellent disease control [1,2,3]. Recent practical data from experienced centers and international registries have shown both sustained treatment benefits and fairly good long-term safety with no major increase in serious infections, cancer, or mortality after more than 5 years of treatment [4,5,6].

Although discontinuation of biologics is associated with a risk of relapse, it may be advisable for several reasons, including a desire expressed by patients to reduce the treatment burden, treatment costs that cannot be covered, the occurrence of adverse events, and specific circumstances such as pregnancy. Time to relapse following treatment discontinuation appears to be very short with some therapeutic agents [7]. Most registration programs for new dermatology drugs include studies with a treatment withdrawal period to assess the potential for rebound effects and to determine the time to psoriasis relapse. The treatment withdrawal period can also be used to assess the effectiveness of treatment reintroduction.

The analysis of the dynamics of psoriasis relapse during the treatment withdrawal period provides relevant information regarding the potential usage of the drug in the clinic.

This systematic review aims to estimate the psoriasis relapse time after systemic drug discontinuation in psoriasis.

2 Material and Methods

The systematic review was prospectively registered at the PROSPERO register under number CRD42021252182.

2.1 Inclusion Criteria

We included randomized controlled trial studies that investigated the time to relapse following discontinuation of systemic treatment with fumarates, acitretin, methotrexate, cyclosporine, infliximab, etanercept, adalimumab, certolizumab pegol, ustekinumab, secukinumab, ixekizumab, brodalumab, bimekizumab, guselkumab, tildrakizumab, risankizumab, tofacitinib, or apremilast in patients with plaque psoriasis.

2.2 Literature Search

For the literature search, we have used the recent Cochrane skin group systematic review from November 2019 [1]. From the 140 studies included, we selected those that included a withdrawal period. A supplementary literature search from the earliest available date to April 2021 was performed in the PubMed, Cochrane Library, and Embase databases using the keywords ("Psoriasis"[Mesh] OR "Plaque psoriasis"[tiab]) AND ("Fumarates"[Mesh] OR "acid fumaric"[tiab] OR "Acitretin"[Mesh] OR "Acitretin"[tiab] OR "Methotrexate"[Mesh] OR "Methotrexate"[tiab] OR "Cyclosporine"[Mesh] OR "Cyclosporine"[tiab] OR "Ciclosporine"[tiab] OR "Infliximab"[Mesh] OR "Infliximab"[tiab] OR "Etanercept"[Mesh] OR "Etanercept"[tiab] OR "Adalimumab"[Mesh] OR "Adalimumab"[tiab] OR "Certolizumab Pegol"[Mesh] OR "Certolizumab Pegol"[tiab] OR "Ustekinumab"[Mesh] OR "Ustekinumab"[tiab] OR "Secukinumab"[Mesh] OR "Secukinumab"[tiab] OR "Ixekizumab"[Mesh] OR "Ixekizumab"[tiab] OR "Brodalumab"[Mesh] OR "Brodalumab"[tiab] OR “Bimekizumab”[Mesh] OR “Bimekizumab”[tiab] OR "Guselkumab"[Mesh] OR "Guselkumab"[tiab] OR "Tildrakizumab"[Mesh] OR "Tildrakizumab"[tiab] OR "Risankizumab"[Mesh] OR "Risankizumab"[tiab] OR "Tofacitinib"[Mesh] OR "Tofacitinib"[tiab] OR "Apremilast"[Mesh] OR "Apremilast"[tiab]) AND ("Drug Tapering"[Mesh] OR Recurrence[Mesh] OR Recurrence[tiab] OR "Withdrawal"[tiab] OR "Withdrawn"[tiab] OR "Discontinue"[tiab] OR "Discontinuation"[tiab] or "Stop"[tiab] OR "Stopped"[tiab] OR "Maintenance"[tiab] OR "Relapse"[tiab] OR "Interruption"[tiab]) AND ("Randomized Controlled Trial"[Publication Type] OR "Randomized"[tiab]) (see the “Supplementary Information” in the electronic supplementary material). The reference lists of included articles were also screened manually for additional studies. Moreover, we contacted individual pharmaceutical companies to provide data missing from the published studies on the time of relapse post-drug withdrawal.

2.3 Data Extraction

The authors (Marie Masson Regnault, Jason Shourick, and Carle Paul) independently searched the databases for potentially relevant article titles and abstracts based on the inclusion criteria. Full text articles were then retrieved. We used the modified Cochrane Collaboration tool to assess risk of bias for randomized controlled trials. Bias was assessed as a judgment (high, low, or unclear) for individual elements from five domains (selection, performance, detection, attrition, and reporting bias) (see the “Supplementary Information” in the electronic supplementary material). The authors were independently involved in all stages of study selection and data extraction in terms of the number of patients enrolled in all of the studies, the number of patients in the withdrawal period, the criteria used for treatment withdrawal, the criteria defining relapse, post-withdrawal follow-up time, and the time to relapse.

3 Results

3.1 Article Selection

Overall, 30 original articles published before April 2021 were included in this review, but several papers report on the same study. Thus, we included 25 studies that report on 27 drug data points about time to relapse (Fig. 1). They are presented in Table 1. No randomized controlled trial was found investigating time to relapse with fumarates or acitretin. Four studies focused on conventional systemic treatments with methotrexate and/or cyclosporine [8,9,10,11], nine focused on tumor necrosis factor (TNF) antagonists [12,13,14,15,16,17,18,19,20], six focused on interleukin-17 (IL-17) antagonists [21,22,23,24,25,26,27, 37], six focused on IL-12/23 or IL-23 antagonists [15, 28,29,30,31,32,33,34], and one study each focused on tofacitinib [35] and apremilast [36]. The number of patients enrolled in these withdrawal-period studies varied from 14 to 917 (mean ± standard deviation, 217.75 ± 191.7; median [interquartile range (IQR)] 186 [84–356.5]).

Fig. 1
figure 1

Article selection

Table 1 Characteristics of the treatment withdrawal period and relapse reported in the studies

3.2 Therapeutic Success Criteria Allowing Patients to Participate in the Withdrawal Period

The therapeutic success criteria determining inclusion of patients in the withdrawal period varied across studies. In three studies, patients had to attain Psoriasis Area Severity Index (PASI) 50 [12] or Physician’s Global Assessment (PGA) ≤ 2 [13, 14]. Patients had to achieve PASI 75 (± PGA 0–1), PASI 90 (± PGA 0–1), or only PGA 0–1 in 15 [8, 11, 15, 16, 18, 20,21,22,23, 26, 28, 30, 32, 35, 36], four [9, 24, 33, 37], and three studies [10, 17, 29], respectively. The therapeutic success criteria became more stringent in recent studies. Indeed, the achievement of PASI 90 or PGA 0–1 was used as a primary end point for the first time with the ixekizumab study published by Gordon et al. in 2016 and thereafter for the brodalumab, guselkumab, risankizumab, and bimekizumab studies [24]. The onset of the withdrawal period varied from 4 to 40 weeks after the beginning of treatment.

3.3 Relapse-Defining Criteria

Similarly, heterogenous criteria were used across different studies to define psoriasis relapse. Relapse was sometimes defined by multiple criteria in the same study (Table 1). The definition of relapse became more stringent in the most recent studies evaluating biological agents.

The duration of follow-up after treatment discontinuation varied from 8 to 108 weeks [mean 33.1 ± 21.58—median 30 (10–20)].

3.4 Time to Relapse

Given the variability in the relapse evaluation criteria, we have computed the time to relapse for each molecule under consideration, ranging from the longest to the shortest time to relapse (Fig. 2a, b). In most cases, the comparison between drugs was indirect. The time to loss of PASI 75 was compared between etanercept and ustekinumab in one study [15].

Fig. 2
figure 2

a Median time to relapse (in weeks) depending on the treatment, with less stringent criteria defining relapse, e.g., loss of PASI 75, loss of PASI 50, 50% loss of PASI maximum improvement, loss of PGA ≤ 2, or when patients felt they needed to resume systemic treatment. b Median time to relapse (in weeks) depending on the treatment, with stringent criteria defining relapse, e.g., loss of PASI 100, loss of PASI 90, or loss of PGA 0 or 1. Imp improvement, PASI Psoriasis Area Severity Index, PGA Physician’s Global Assessment

3.4.1 Oral Systemic Treatment

For patients treated with oral cyclosporine, four studies evaluated time to relapse in 25, 14, 20, and 192 patients, respectively [8,9,10,11]. Each study used a different relapse definition. Relapse was defined as the moment when patients felt a new course of cyclosporine was required, as a 2-point reduction in a 7-point severity scale, or as a 50% loss of maximum PASI improvement. The median length of time to the perceived need for a new course of cyclosporine treatment varied from 9.5 to 16 weeks. The median time to loss of 50% of maximum PASI improvement was about 4 weeks. For methotrexate patients in a small study (n = 17), relapse was defined as a 50% loss of maximum PASI improvement or when the patients felt treatment was needed [9]. Similarly to cyclosporine, the median time to loss of 50% of maximum PASI improvement was 4 weeks (Fig. 2a).

3.4.2 TNF Antagonists

Four studies focused on etanercept [12,13,14,15], two studies each focused on adalimumab [16, 17] and infliximab [18, 19], and one study focused on certolizumab pegol [20]. A higher number of patients were included in the withdrawal period in the TNF antagonist studies compared to conventional systemic treatment, ranging from 99 to 917 patients [mean 330.2 ± 257.8; median 240 (149–409)]. The duration of follow-up ranged from 8 to 36 weeks. Time to relapse was defined as loss of PGA ≤ 2 in five studies [13,14,15, 17, 18], an increase of more than 6 points in the PASI score in one study [16], a loss of PASI 50 in two studies [12, 16], a 50% reduction in maximum PASI response in two studies [12, 20], and a loss of PASI 75 in three studies [12, 18, 19]. The median time to loss of PGA ≤ 2 varied from 4.7 to 7.3 weeks for etanercept and was estimated to be about 20.1 weeks for adalimumab. The median time to loss of PASI 50 for etanercept and adalimumab was estimated to be about 13 and 23 weeks, respectively. The median time to loss of PASI 75 was about 8.1 weeks for etanercept and approximately 15 weeks for infliximab. The median time to loss of 50% loss of maximum PASI improvement was about 12 weeks for etanercept and 22 weeks and 20 weeks for certolizumab pegol 200 mg and 400 mg, respectively (Fig. 2a).

3.4.3 IL-17 Antagonists

Three studies were identified for secukinumab [21,22,23], two for ixekizumab [24, 25], two for brodalumab [26, 27], and one for bimekizumab [37]. The number of patients included varied from 67 to 423 patients [mean 171.6 ± 134.1; median 132 (84–188.5)]. The time to relapse was reported as a loss of PGA ≤ 2 in four studies [24,25,26,27], a one-third loss of maximum PASI improvement in one study [21], a loss of 50% of maximum PASI improvement in one study [23], a loss of PASI 75 for three studies [22, 24, 25, 37], a loss of PASI 90 in four studies [23, 24, 26, 37], a loss of PASI 100 in three studies [24, 26, 37], and a loss of PGA 0–1 in one study [26]. The median time to loss of PGA ≤ 2 was about 20 weeks for ixekizumab and 9 weeks for brodalumab. The median time to loss of PASI 75 varied from 20 to 24 weeks for secukinumab 300 mg. It was about 20 weeks for ixekizumab and estimated at 32 weeks for bimekizumab. The median time to loss of PGA 0–1 for brodalumab was estimated to be between 6 and 7 weeks. For brodalumab, ixekizumab, secukinumab 300 mg, and bimekizumab, the median time to loss of PASI 90 was estimated to be 7 weeks, 16 weeks, 16 weeks, and 24 weeks, respectively. For brodalumab, ixekizumab, and bimekizumab, the median time to loss of PASI 100 was estimated to be 6 weeks, 12 weeks, and 18 weeks, respectively (Fig. 2a, b).

3.4.4 IL-12/23 and IL-23 Antagonists

We included two studies for ustekinumab [15, 28], two for guselkumab [33, 34], one for risankizumab [29], and three for tildrakizumab [30,31,32]. The number of patients included in the withdrawal period ranged from 113 to 399 [mean 243.2 ± 105.8; median 229 (176.3–305)]. Time to relapse was reported as a loss of 50% maximum PASI improvement in nine studies [9, 12, 20, 23, 28, 30,31,32,33,34,35], a loss of PASI 75 in five studies [28, 30,31,32,33], a loss of PASI 90 in six studies [28, 29, 31,32,33,34], a loss of PGA 0–1 in three studies [26, 27, 29,30,31], and a loss of PASI 100 in two studies [29, 33]. The median time to loss of PASI 75 was about 15 weeks for ustekinumab compared to over 28 weeks for guselkumab and about 32.3 for tildrakizumab (100 mg). The median time to loss of PASI 90 was estimated at 20 weeks for ustekinumab, between 23 and 24 weeks for guselkumab, 27.9 weeks for tildrakizumab (100 mg), and 42 weeks for risankizumab. The median time to loss of PGA 0–1 was estimated to be 28 weeks for guselkumab, 28 weeks for tildrakizumab (100 mg), and 44 weeks for risankizumab (Fig. 2a, b).

3.4.5 Small Molecule Inhibitors

We found one tofacitinib study including 291 patients in the withdrawal period [35]. Patients who obtained PASI 75 and PGA 0–1 responses were selected to enter the withdrawal period. The relapse was defined as a 50% loss in maximum PASI improvement during initial treatment or a loss of a PASI 75 response. Median time to a 50% loss of maximum PASI response was 16.1 weeks. Median time to a loss of PASI 75 response was 8 weeks (Fig. 2a). We found one study for apremilast [36]. The relapse was defined as a loss of PASI 75. The median time to initial loss of PASI 75 after re-randomization at week 32 was 5.1 weeks (95% confidence interval 4.1–8.1) (Fig. 2a).

3.5 Drug Half-Lives

The mean terminal elimination half-life of different drugs was collected from product information or public assessment reports available on the European Medicines Agency’s (EMA) website. These are presented from the longest to the shortest maintenance response we determined in the previous subsection (Fig. 3).

Fig. 3
figure 3

Mean terminal elimination half-life (in days) of evaluated drugs. Data were extracted from the product information or public assessment report for each drug (available at Drugs are ranked from left to right from the longest to the shortest time to relapse.

4 Discussion

Systemic treatment is used as induction and maintenance therapy in patients with moderate to severe psoriasis. In a real-life setting, discontinuation of treatment may be triggered by life events such as pregnancy or concomitant disease. It may also be motivated by a patient’s desire to have drug-free intervals or to reduce the treatment burden.

Based on indirect comparison, overall, we noted a longer period of maintained response after discontinuation of biological agents as compared with oral systemic treatment. Moreover, treatment with IL-12/23 and IL-23 antagonists showed a tendency for a longer time to relapse compared to IL-17 and TNF antagonists. It should be noted that due to clinical trial constraints, the drug discontinuation was performed relatively early. It remains to be determined whether discontinuation after a long period in remission would produce the same results. Future studies are needed to investigate whether drug discontinuation after sustained complete remission of psoriasis can be associated with long-term psoriasis clearance in some patients. The time to relapse after treatment discontinuation is particularly interesting in clinical settings where the use of systemic agents may be detrimental. A recent example is the worldwide coronavirus disease 2019 (COVID-19) pandemic, during which practitioners were initially concerned about maintaining treatment for patients until reassuring data were published [38]. In 2010, Kamaria et al. produced a non-systematic review comparing biologics and their respective times to relapse [39]. They showed that alefacept offered the longest clinical benefit, followed by ustekinumab, infliximab, adalimumab, etanercept, and lastly, efalizumab. More recently, Wang et al. reported a short non-systematic review on time to relapse after biologic treatment withdrawal including observational studies. They suggested that IL-23 inhibitors are associated with a longer time to relapse after treatment withdrawal than TNF inhibitors [40].

Observational studies provide additional information about real-life psoriasis relapse following the discontinuation of biologic agents in psoriasis. Bellinato et al. reported, in 53 patients treated with etanercept and achieving PASI 0/1, that 33 (62%) relapsed within 6 months after drug discontinuation. The median time to relapse (defined as PASI score variation ≥ 5) was 184 days (~ 26 weeks) [41]. Giunta et al. reported a median time to loss of PASI 50 of 182 days (range 84–308 days) (~ 26 weeks in patients experiencing a PASI 75 response to infliximab) [42]. For IL-17 antagonists, we reported a 46-day median time to psoriasis relapse for patients stopping brodalumab [7]. Umezawa et al. reported, in PASI 75 responders to ixekizumab, that approximately half of the patients relapsed (loss of PASI 50) within 5 months after drug discontinuation. [43]. Rivera et al. reported a median time from the last dose of biologics to the initiation of new systemic treatment (relapse) of 192.5 days (IQR 107–308 days) (~ 27 weeks) for patients treated with secukinumab [44]. In an 8-year multicenter retrospective study, Chiu et al. reported in 202 patients responding to ustekinumab a median time to relapse (loss of 50% of the PASI maximum improvement) of 6.7 ± 4.1 months. Possible clinical predictors for longer relapse times were being biologic-naive, having a shorter psoriasis duration, absence of arthritis or chronic kidney disease, absence of family history of psoriasis, and higher and more rapid response to treatment [45]. Rivera et al. reported a median time to relapse of 282 days in patients treated with guselkumab [44]. The conclusions about these findings are limited by the very large and variable definitions used for defining relapse. Down-titration and discontinuation strategies of a biological treatment are also a topic of research in other inflammatory diseases, such as rheumatoid arthritis. It has been shown that fixed-dose reduction of TNF inhibitors is associated with maintenance of efficacy, whereas drug withdrawal leads to disease relapse [46].

One important factor to consider when evaluating continued drug effects after treatment discontinuation is the persistence of the drug in the body. Based on pharmacokinetic data, maintained response following drug discontinuation appears to partially correlate with the mean drug half-lives. Overall, the longest time to relapse was recorded with drugs with the longest half-lives. The difference in efficacy maintenance between drugs cannot be solely explained by drug half-lives. Other factors may play a role, such as molecular target and site of action in the psoriasis inflammation cascade.

The main study limitation is the scarcity of published data currently available on clinical study withdrawal periods and the considerable variability in the criteria used to define optimal clinical response before treatment discontinuation and relapse during the withdrawal period. Relapse after the discontinuation of oral systemic agents and TNF antagonists have mostly been evaluated using less-stringent measures. Only limited data were available for traditional systemic treatments such as methotrexate and cyclosporine, and no data exists for acitretin or fumarates. A consensus approach to the definition of psoriasis relapse in the treatment withdrawal period of clinical studies is needed, taking both patient and physician perspectives into account. The absence of access to individual patient data does not allow assessment of the role of demographic factors [e.g., body mass index (BMI), age, sex] or clinical variables such as baseline disease severity or age of onset of psoriasis. The effects of discontinuation of systemic therapy in psoriasis patients are unclear. In the case of molecules with long-term efficacy after discontinuation, evidence suggests that it may be possible to increase the interval between doses in some patients for IL-23 antagonists. Further high-quality research is needed in this area, with particular emphasis on double-blind comparative randomized controlled trials in which biologic therapy or immunosuppressants are withdrawn, as well as registry data on the real-life efficacy and safety of biologic discontinuation in patients.

5 Conclusion

The data from this systematic review suggest that biological agents are associated with improved maintenance of the effect and longer time to relapse compared to oral systemic agents. Among biologicals and when considering the most stringent criteria, it seems that IL-23 antagonists are associated with a longer time to relapse post-withdrawal compared to other psoriasis treatments. These findings may have clinical consequences on the selection of systemic agents in certain situations. Indeed, when the need for systemic treatment discontinuation can be anticipated, such as for scheduled surgery, planned pregnancy, or long-term travel to countries with limited access to healthcare resources, IL-23 antagonists may be preferred based on their persistent effect.