A living systematic review and network meta-analysis by the Cochrane Library has compiled phase II–IV RCTs of systemic therapies in adult patients with moderate-to-severe plaque psoriasis through to October 2021 (9). Eligible studies from this living review were included for analysis. The data was supplemented by searching Embase (Ovid), MEDLINE (Ovid), and the Cochrane CENTRAL Register of Controlled Trials up to January 1st, 2023. The full search strategy is outlined in Supplementary Table 1 and includes terms to capture psoriasis, systemic therapy, and placebo-controlled trials. References of relevant publications were also screened, and only studies published in the English language were included.

Studies were included for analysis using the following criteria: placebo-controlled phase II, III, or IV induction or maintenance clinical trials of patients aged ≥18 with moderate-to-severe plaque psoriasis; evaluation of conventional systemic anti-psoriatic agent [defined as methotrexate, cyclosporine, oral retinoid, fumaric acid ester, biologic, and/or oral small molecule (apremilast or deucravacitinib)] versus placebo; and published frequency and nature of AEs (including any AE, SAE, AE requiring treatment discontinuation, infections, and/or injection- or infusion-related AE) in both treatment and placebo groups. Phase I clinical trials were excluded given substantial methodological differences compared to phase II to IV studies.

All citations were independently reviewed by two separate investigators (BM and Y-JP) using the above predefined inclusion criteria. Studies were screened by title and abstract followed by full text review. Disagreements were settled by a third author (PM). All screening was performed using Covidence Systematic Review software (Veritas Health Innovation, Melbourne, Australia).

The primary outcome was the pooled proportion of patients experiencing any AE in the placebo arm, and associated RD between systemic therapy and placebo. Secondary outcomes included pooled RDs in SAEs, AEs requiring treatment discontinuation, infections, and infusion- or injection-related AEs between treatment and placebo groups. All AEs were defined and reported by the original study authors. For trials testing multiple interventions, the proportion of patients with each outcome was pooled by treatment class. Multiple doses of systemic therapy were pooled if applicable. Data only over the initial placebo-controlled portion of trials were included.

Trial features that were extracted included: study design and setting (phase, number of centers, duration of follow-up); psoriasis severity criteria; number of patients randomized to systemic therapy and placebo; and incidence and nature of AEs in both intervention and treatment arms. The Cochrane Collaboration Risk of Bias tool version 2.0 was used to assess the methodological quality of included trials (16).

The proportion of patients experiencing the primary and secondary outcomes in each of the placebo and active treatment arms were pooled using a random effects model to account for between- and within-study heterogeneity. The Freeman-Tukey double arcsine transformation was used to compute 95% confidence intervals (CI) using the score statistic and exact binomial method. The pooled RD between placebo and intervention arms stratified by medication class (biologic versus non-biologic) were calculated using a restricted maximum likelihood random effects model with 95% CIs. Statistical heterogeneity was quantified using Cochran’s Q and I² statistics, and interpreted based on Cochrane recommendations (I² = 30–60% representing moderate, 50–90% substantial, and 75–100% considerable heterogeneity). Univariate meta-regression was used to explore potential causes of heterogeneity using the variables of publication year, trial phase, multinational versus single country study, number of trial centers, and medication class. Publication bias was assessed using funnel plots and Egger’s test. All analyses were performed in Review Manager 5.4 and Stata 17.0 using the metaprop program (StataCorp LLC, College Station, TX).

The final analysis included 103 unique RCTs representing 92 comparisons of biologic therapies and 38 comparisons of non-biologic treatments versus placebo (Figure 1 and Supplementary Appendix 1), enrolling a total of 30,249 patients randomized to systemic therapy (25,067 biologic, 82.9%) and 12,940 patients randomized to placebo. Ninety-six trials had initial placebo-controlled periods of 16 weeks or shorter. Four clinical trials comparing acitretin versus placebo were excluded from the analysis due to an inability to locate and confirm the primary outcomes (17–20). Any AE was reported in 96 comparisons (86 biologic and 10 non-biologic), SAEs in 110 comparisons (91 biologic and 19 non-biologic), AE leading to discontinuation of therapy in 107 comparisons (87 biologic and 20 non-biologic), and infectious AE in 101 comparisons (89 biologic and 12 non-biologic). A total of 52 comparisons of biologic agents reported either injection or infusion-related AEs. There was good agreement between reviewers on final studies for inclusion (Cohen’s Kappa 0.58, 93.8% agreement). Most trials were considered at low risk of randomization, missing data, and reporting bias (Supplementary Tables 2, 3).

The pooled RD for any AE between systemic therapy and placebo is summarized in Figure 2 and Table 1. A total of 49.8% [95% CI: 47.1–52.4%] of patients randomized to placebo experienced an AE, compared to 57.1% [95% CI: 54.7–59.5%] in systemic therapy groups, resulting in a RD of 6.7% [95% CI: 4.6–8.9%, p < 0.00001] with considerable overall heterogeneity (I² = 75%). This RD was observed in subgroup analyses for both biologic [RD 5.3% (95% CI: 3.0–7.5%), p < 0.00001, I² = 71%] and non-biologic therapies [RD 12.6% (95% CI 7.3–18.0%), p < 0.00001, I² = 81%] compared to placebo. There was no evidence of publication bias (Supplementary Figure 1, Egger p-value = 0.45).

The pooled proportion of patients receiving placebo who experienced an SAE was 1.4% [95% CI: 1.2–1.7%], and 1.9% [95% CI: 1.5–2.3%] of placebo patients discontinued therapy due to an AE. No statistically significant risk difference between patients who received systemic therapy and placebo for SAEs [RD 0.3% (95% CI 0.0–0.6%), p = 0.06, Figure 3] or AEs necessitating discontinuation of therapy [RD 0.2% (95% CI −0.2–0.6%), p = 0.28, Figure 4] was observed. In subgroup analysis, 6.5% (95% CI: 4.1–9.3%) of patients exposed to non-biologic agents experienced AEs requiring medication discontinuation compared to 3.6% [95% CI: 2.7–4.7%] of placebo patients (RD 2.9% [95% CI: 0.5–5.2%], p = 0.02) with considerable heterogeneity (I² = 78%). Subclassification by year of publication, trial phase, multinational or single country, number of centers, and drug class in meta-regression resolved heterogeneity in SAEs (I² = 14.75%), but only partially explained heterogeneity in AE leading to discontinuation (I² = 49.50%).

There was a higher risk of infections in patients receiving systemic therapy compared to placebo [RD 4.3% (95% CI 3.0–5.7%), p < 0.00001] that persisted in subgroup analyses for biologic agents [RD 4.3% (95% CI 2.8–5.8%), p < 0.00001, I² = 75%, Figure 5] and non-biologic agents [RD 4.1% (95% CI: 1.5–6.8%), p = 0.002, I² = 23%].

Injection- and infusion-related AEs are summarized in Supplementary Figure 2. There was a significant increase in the risk of injection or infusion-related AEs in the systemic therapy groups compared to placebo [RD 3.4% (95% CI 1.9–4.9%), p < 0.00001], although with significant heterogeneity (I² = 89%). A total of 1.7% (95% CI: 1.0–2.5%) of placebo-treated patients experienced an injection or infusion-related AE.

Randomized controlled trials (RCTs) of biologic and non-biologic agents for psoriasis have consistently identified non-specific medication-exposure related AEs (ex. nausea and headache) prone to nocebo effects (9). These effects have been well characterized in numerous drug classes, including HMG-CoA reductase inhibitors, (21) anti-depressants, (22) anti-epileptics, (23) and biologic therapies for other indications (24). However, there has been limited analysis of its role in systemic agents used for dermatologic indications. The pathogenesis of nocebo responses is complex and multifactorial, encompassing negative patient expectations secondary to perceived sensitivity to therapy, prior treatment experiences, and patient-provider therapeutic relationships (25–28). Furthermore, social conditioning from observed responses in others plays a key role, which is especially relevant given modern mass and social media-facilitated distribution of patient experiences with negative side effects and AEs (29). Therefore, patients with dermatologic conditions may be at high risk for nocebo effects because dermatologic diseases are generally highly visible, distressing, and frequently subject to both personal and peer judgment (i.e., stigmatization) that can predispose to negative interpretations of treatment-related events; and have a chronic, relapsing course such that patients may have negative expectations from failing multiple prior combination treatment regimens consisting of both systemic and topical agents.

Nocebo effects have important implications for RCT interpretation. First, high rates of AEs in patients randomized to placebo should be used to contextualize the overall safety profile of novel therapies in RCTs. Second, RCTs are essential for determining the efficacy of novel treatments. Still, they are generally under-powered for evaluating safety, especially for rare or serious AEs that require longer follow-up durations and greater patient exposure time, which is difficult to accommodate in most phase II and III induction and 1-year maintenance studies (30, 31). Third, nocebo effects in intervention groups may artificially increase rates of reported AEs and potentially mask the identification of true treatment-emergent AEs. Fourth, nocebo effects may lead to discontinuation of therapy and trial withdrawal, which can confound both evaluations of efficacy and safety if withdrawals occur differentially in treatment and placebo groups (32, 33). This has previously been observed in trials of patients switching from bio-originator to biosimilar TNF antagonists, despite numerous studies demonstrating bio-equivalence and non-inferiority (26). Reassuringly, we identified similar rates of treatment discontinuation between treatment and placebo due to AEs in psoriasis trials. Taken together, these findings highlight the importance of post-marketing drug registries, open-label trial extensions, and integrated safety analyses characterizing long-term safety outcomes.

Recognizing and minimizing nocebo effects in clinical practice may improve patient outcomes and enhance treatment persistence. While it is critical to ensure that all patients starting systemic therapy are informed of the risks and benefits of treatment, several strategies have been proposed to minimize nocebo effects. These include positive framing of side effect profiles, explicit disclosure of possible nocebo effects, standardized approaches to questioning for and measuring patient-reported AEs, and authorized concealment of limited disclosure of potentially rare or irrelevant AEs (28, 34–36). However, our findings need to be cautiously generalized to real-world practice given that clinical trials often select for an overall healthier patient population, whereas more comorbid patients may be using multiple concomitant therapies and be intrinsically at higher risk of AEs; and clinical constraints may alter the informed consent process and presentation of potential side effects compared to a controlled trial setting. The treatment context of an RCT itself may influence patient reporting of AEs because the processes of randomization, informed consent, and blinding have all been linked to nocebo effects (28, 37–39). For example, inclusion of possible gastrointestinal upset in written consent forms for unstable angina therapy was shown to increase the proportion of patients withdrawing from the study due to subjective, minor gastrointestinal symptoms by sixfold (39).

Our study has several important strengths. This systematic review and meta-analysis uniquely assesses the pooled proportion of patients experiencing AEs and associated RD between placebo and both biologic and non-biologic therapies in over 30,000 psoriasis patients. However, our study has some key limitations. First, alternative explanations for the high rate of AEs observed in placebo groups should also consider the accumulation of psoriatic complications from the natural history of progressive, untreated disease; potential misattribution of symptoms from related, comorbid psoriatic conditions such as psoriatic arthritis; and potential effects of concomitant topical or systemic therapies for either psoriasis or an associated condition. Second, there was significant heterogeneity between studies when assessing for any AE that was not fully explained in meta-regression. This may be a consequence of differences in defining and reporting AEs, patient characteristics (such as psoriatic involvement of special sites, concomitant psoriatic complications), and/or intervention differences between sub-therapy classes. For example, oral non-biologic medications, like cyclosporine and acitretin, have different mechanisms of action, resulting in potentially distinct side effect profiles.

In conclusion, we performed a comprehensive systematic review and meta-analysis of systemic therapies for moderate-to-severe plaque psoriasis. Nearly half of all patients randomized to placebo experienced AEs. Our evaluation reveals the necessity of considering nocebo effects to account for these findings. We did not identify any significant overall RD in either serious AE or AE leading to discontinuation of therapy between systemic therapy and placebo. These outcomes inform the interpretation of RCT data and influence clinician-patient communication.