Risk of Paradoxical Eczema in Patients Receiving Biologics for Psoriasis

This cohort study examines which biologic drug classes and factors are associated with risk of paradoxical eczema in patients with psoriasis treated with biologics.


W
hile biologics targeting tumor necrosis factor (TNF), interleukin (IL) 12/23, IL-17, and IL-23 are highly effective treatments for plaque psoriasis, they are associated with cutaneous adverse events, such as paradoxical psoriasis, 1 cutaneous lupus, and granulomatous disorders. 2ome patients with psoriasis develop paradoxical eczema, an atopic dermatitis (AD) phenotype, during biologic exposure 3 because these dermatoses are genetically and immunologically divergent and rarely occur together [4][5][6][7] ; 1 meta-analysis identified a 2% prevalence of AD in those with psoriasis. 8Furthermore, there have been reports of psoriasis or inflammatory arthritis developing secondary to IL-4/13 inhibitors used for AD. 9,10n addition to the direct impact of paradoxical eczema, this could result in treatment discontinuation or use of concomitant immunosuppressants. 3It is unclear whether paradoxical eczema risk varies by biologic class or other clinical features.The British Association of Dermatologists Biologics and Immunomodulators Register (BADBIR) has recruited more than 20 000 patients with psoriasis from 168 centers in the UK and Ireland. 11We used BADBIR to undertake a prospective cohort study to assess (1) the overall and biologic class-specific incidence of paradoxical eczema, (2) whether risk of paradoxical eczema differs between TNF inhibitors and other biologic classes, and (3) the demographic and clinical factors associated with paradoxical eczema.

Methods
The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline was followed for the reporting of this cohort study.This work was completed under the ethics approvals of BADBIR.BADBIR was approved in March 2007 by the National Health Service Research Ethics Committee North West England.Written informed consent was obtained from all participants before enrollment.

Study Participants
Using BADBIR data (from September 2007 to December 2022), we included adults aged 18 years or older with plaque psoriasis exposed to 1 of the following biologics (biosimilars treated the same as originators): TNF inhibitors (adalimumab, certolizumab pegol, etanercept, or infliximab), IL-17 inhibitors (bimekizumab, brodalumab, ixekizumab, or secukinumab), IL-12/23 inhibitors (ustekinumab), or IL-23 inhibitors (guselkumab, risankizumab, or tildrakizumab).As previously described, 12 data were collected contemporaneously at baseline, every 6 months for the first 3 years, and annually thereafter.Data collected included systemic therapy start and stop dates, baseline comorbidities, adverse event data, and baseline demographics.Ethnicity was included alongside other variables in this study to explore its association with paradoxical eczema.Ethnicity was classified by study participants.Options, defined on the baseline registration questionnaire, included Black, Chinese, South Asian, White, and other (with free text for participants to specify other ethnicities).Exposures with only 1 recorded entry and no follow-up were excluded (Figure).

Cohort Study Design
For the primary analysis, all lines of exposure were included.Exposures contributed follow-up time until 1 of the following occurrences: (1) paradoxical eczema onset, (2) treatment switch or discontinuation, (3) last documented follow-up, or (4) death.An exposure was considered continuous if there was a treatment break of 90 days or less before restarting the same biologic.Exposures with concomitant conventional systemic or small-molecule therapies were included.A risk window of 90 days was used for the primary analysis; if a paradoxical eczema event occurred within 90 days of switching to another biologic, it was counted as an event for both biologic agents.The final sample size was determined by the number of eligible exposures.

Paradoxical Eczema Event Definition
Paradoxical eczema events were identified by reviewing adverse event data, including clinical descriptions from the participant's case records.Events containing the terms eczema, eczematized, eczematous, atopy, atopic, or dermatitis were screened.Potential events were reviewed by 2 researchers (A.A.-J.and R.B.W.) and included if they were described as eczema, atopic eczema, atopic dermatitis, or psoriasiform eczema or eczematized psoriasis.Events were excluded if they were noneczema events (eg, perioral dermatitis), alternative phenotypes (eg, contact dermatitis), or duplicate records.

Statistical Analyses
Statistical analyses were undertaken in Stata, version 14.2 (StataCorp LLC). 13Crude incidence rates of paradoxical eczema and descriptive statistics for paradoxical eczema events (the eMethods in Supplement 1 give more detail) are provided.For the primary analysis, we used a Cox proportional hazards regression survival analysis model to compare the probability of paradoxical eczema in those receiving IL-17 inhibitors, IL-12/23 inhibitors, or IL-23 inhibitors compared with TNF inhibitors. 14This was repeated for individual biologics vs adalimumab.Because all lines of exposure were included, we used the vce(cluster id) command in Stata to adjust for the correlation of observations coming from the same participant.We tested for the validity of the proportional hazards assumption for Cox modeling using Schoenfeld residuals.

Key Points
Question What factors are associated with paradoxical eczema occurring in patients with psoriasis treated with biologics?Findings In this cohort study of 24 997 biologic exposures in 13 699 patients with psoriasis, risk of paradoxical eczema was lowest in patients receiving interleukin 23 inhibitors compared with other biologic classes.Increasing age, history of atopic dermatitis, and history of hay fever were associated with higher risk of paradoxical eczema; risk was lower in males.

Meaning
The findings suggest that interleukin 23 inhibitors could be considered in patients with psoriasis with factors associated with paradoxical eczema.
To adjust for confounders, we generated biologic class propensity scores for inverse-probability treatment weighting for each observation. 15,16After weighting, the covariate balance between treatment arms was equivalent (eTable 1 in Supplement 1).To maximize precision of the effect estimate and minimize bias, the propensity score model included covariates suspected to be associated with the outcome but not the exposure as well as potential confounders (the eMethods in Supplement 1 gives for more detail on confounder selection). 17or our primary analysis, missing data for included covariates was 0.21% or less (eTable 2 in Supplement 1).The 20 participants with missing ethnicity data were allocated to the other category.Participants with missing data on previous alternative psoriasis phenotypes were presumed not to have had the phenotype.For the first-line exposure sensitivity analysis, due to the proportion of missing data in the smoking and alcohol fields (eTable 3 in Supplement 1), missing data were handled with multiple imputation of 20 data sets. 18Propensity scores were calculated separately in each data set, and effect estimates were combined using the Rubin rules. 19o explore the association between demographic and clinical variables and paradoxical eczema, we repeated the propensity score-weighted survival analysis but included all covariates in 1 model and reported their adjusted coefficients.These variables were the same as the list of potential confounders.To further explore risk of paradoxical eczema by age, we repeated the survival analysis using age categories.We additionally undertook several sensitivity and post hoc analyses (eMethods in Supplement 1).Two-sided P < .05 was used for all analyses.

Incidence Rates
A total of 265 paradoxical eczema events were attributed to 273 biologic exposures (1% of total); 8 events occurred within the 90-day risk window of a previous biologic exposure.Adjusted incidence rates were 1.22 per 100 000 person-years for IL-17 inhibitors, 0.94 per 100 000 person-years for TNF inhibitors, 0.80 per 100 000 person-years for IL-12/23 inhibitors, and 0.56 per 100 000 person-years for IL-23 inhibitors (Table 2).Drug-specific incidence rates are presented in Table 2.

Sensitivity Analyses
A repeated analysis with a drug-exposure risk window of 0 days demonstrated similar results to the 90-day risk window primary analysis (eTable 7 in Supplement 1).Due to the potential impact of unmeasured time-varying confounders, we repeated the survival analysis restricted to only the first biologic exposure for each participant.This analysis also assessed paradoxical eczema risk by baseline smoking or alcohol consumption status.Of the 11 732 exposures (exposure time, 39 274 patient-years), paradoxical eczema occurred in 141 (1%).The biologic class HRs were similar to those in the primary analysis (eTable 8 in Supplement 1).For individual biologics, a first-line etanercept exposure was associated with lower risk of paradoxical eczema compared with adalimumab (HR, 0.45; 95% CI, 0.22-0.95).The apparent increased risk associated with certolizumab pegol (HR, 5.41; 95% CI, 1.66-17.57)may have been due to a low number of first-line exposures (n = 57) to this drug (eTable 8 in Supplement 1).There was no association with baseline smoking status or alcohol consumption (eTable 8 in Supplement 1).To evaluate whether our findings were specific to the paradoxical eczema phenotype, we repeated the survival analysis but defined treatment failure as onset of other eczema reactions (n = 156), such as contact dermatitis, seborrheic dermatitis, and stasis dermatitis (eTable 9 in Supplement 1).There were no significant differences in risks for these phenotypes between biologics and biologic classes (eTable 10 in Supplement 1).Although there were no significant differences, the HRs were greater for IL-12/23 inhibitor exposures (HR, 1.07; 95% CI, 0.74-1.56)and IL-23 inhibitor exposures (HR, 1.25; 95% CI, 0.65-2.44)than for TNF inhibitor exposures; the opposite was observed for paradoxical eczema.The direction of the coefficients for age, male sex, AD, and hay fever were the same as for paradoxical eczema, although there was no significant difference for male sex and AD for other eczema phenotypes.Post hoc, we found that the median time to all-cause biologic discontinuation following onset of paradoxical eczema (467 days; To assess the impact of cotreatment with nonbiologic systemic therapies, we repeated the model including this as a binary variable and identified no association with paradoxical eczema (HR, 1.26; 95% CI, 0.89-1.79).When nonbiologic systemics were categorized, we observed an increased risk of paradoxical eczema associated with cyclosporine (HR, 3.28; 95% CI, 2.03-5.30)and no association with methotrexate (HR, 0.62; 95% CI, 0.37-1.04)(eTable 11 in Supplement 1).To account for biologic exposure-related covariates that were not included, such as treatment failure, treatment of previous paradoxical eczema events, or occurrence of other adverse events, we repeated analysis for first-line biologic exposures only.The increased risk of paradoxical eczema associated with concomitant cyclosporine remained (HR, 2.10; 95% CI, 1.12-3.91),and there was no association with methotrexate (HR, 0.78; 95% CI, 0.41-1.47).Of 30 patients receiving cyclosporine at biologic initiation who then developed paradoxical eczema, the recorded onset date of paradoxical eczema was after or on the same day as cyclosporine cessation in 21 cases (median, 0.5 days; IQR, −28 to 101 days), with the distribution demonstrating a left-sided skew (eFigure 4 in Supplement 1), unlike the methotrexate cohort (n = 17; median, −6 days; IQR, −330 to 137 days).

Discussion
In this study, while the incidence of paradoxical eczema in biologic-treated patients with psoriasis was low overall, it was highest in those receiving IL-17 inhibitors followed by those receiving TNF inhibitors, those receiving IL-12/23 inhibitors, and those receiving IL-23 inhibitors.Compared with TNF inhibitors, IL-23 inhibitor exposure was associated with significantly lower risk of paradoxical eczema; this result may have been attributable mostly to guselkumab due to the low number of exposures to other IL-23 inhibitors in these data.These findings remained when restricting the analysis to first-line biologic exposures and were specific to this eczema phenotype.Increasing age, female sex, prior AD, and prior hay fever were associated with increased risk of paradoxical eczema.

Interpretation of Findings
To our knowledge, this is the first study to compare paradoxical eczema risk by biologic class.Based on clinical experience and prevalence of eczematous reactions reported in some IL-17 inhibitor clinical trials, [20][21][22][23] we suspected an association between IL-17 inhibitor exposure and paradoxical eczema.While the incidence of paradoxical eczema was numerically highest among IL-17 inhibitor exposures, it was not significantly different from the incidence among TNF inhibitor exposures.The low overall incidence of paradoxical eczema may be reassuring for patients and clinicians, but it is possible that the incidence was underestimated due to underreporting or exclusion of adverse events with insufficient detail.
The mechanisms of paradoxical eczema are unknown.Some authors have speculated that inhibition of TNF or the IL-17/23 axis permits development of T-helper 2 (Th2)-mediated inflammation, which may otherwise be inhibited by Th1/Th17 activity. 3,24Th2-predominant or mixed inflammatory profiles in lesional skin have been described in small case series. 25,26The biological basis for IL-23 inhibitors being associated with the lowest risk of paradoxical eczema is unclear.
Regarding subgroup analyses, the reduced risk of paradoxical eczema associated with guselkumab supports the findings of the primary analysis.Etanercept was associated with a lower risk of paradoxical eczema when the analysis was restricted to first-line exposures, possibly because etanercept was more commonly used as a first-line therapy when BADBIR was established and data entry practices have changed or awareness of this adverse event has increased.
As expected and consistent with other studies, we identified an association between paradoxical eczema and previous AD [27][28][29] and hay fever.This supports an association of genetic factors with atopy, which we previously demonstrated in a genotyped cohort with paradoxical eczema. 30The lack of association with asthma could indicate that certain genetic variants more commonly associated with AD and hay fever rather than asthma, such as FLG variants, 31 play an important role in paradoxical eczema.
To our knowledge, the increased risk of paradoxical eczema in females has not been identified previously.While AD is more common males than in females during childhood, this trend reverses into early adulthood; this finding in our study could reflect this. 32We did not observe an increase in paradoxical eczema risk in South Asian participants compared with White participants, but could not conclude on risk in other ethnic groups due to sample size limitations.Unlike a previous study, 27 we did not identify an increased risk in smokers or those consuming alcohol at baseline.
We were surprised to find an association between cyclosporine use at the time of biologic initiation and paradoxical eczema.Some patients may have had an undocumented active eczema, or an overlapping phenotype, that was unmasked at cyclosporine withdrawal or dose reduction or biologic initiation.Cyclosporine treatment or withdrawal has been previously reported to aggravate or induce relapse in animal autoimmunity models, possibly by suppressing inflammation while allowing antigen-specific priming of T cells, altering Th1/Th2 antagonism, 33 or inactivating regulatory T cells. 34ome patients developed repeated paradoxical eczema events while receiving different biologic classes.This finding may indicate a common immunological mechanism downstream of the drug targets.While there were too few participants for inferential statistical analysis, most patients with multiple events developed their first event while receiving a TNF inhibitor, and a substantial proportion were using cyclosporine at biologic initiation.Assessment of the consequences of specific sequences of therapeutic agents and drug combinations is a challenging but important area for future research.

Strengths and Limitations
Strengths of this study include the large sample size and inclusion of multiple lines of exposure per participant.We included data for all currently available biologics, originating from more than 160 dermatology centers in the UK and Ireland.We minimized bias from our confounders using propensity score weighting.
The main limitation is the small numbers of observations within certain subgroups, such as specific biologic exposures or participants in ethnic minority groups, restricting generalizability of our findings and the interpretation of some subgroup analyses.The small number of paradoxical eczema events resulted in imprecise effect estimates.The reduced risk observed in association with IL-23 inhibitors should be interpreted with caution, as the number of IL-23 inhibitor exposures was low compared with other classes.There was a risk of adverse event misclassification, as ascertainment of adverse events relied on free-text descriptions derived from medical records, which are subjective.It was also not possible to ascertain whether each paradoxical eczema event was associated with the drug or represented natural occurrence of adultonset eczema, but the tendency for events to occur close to biologic initiation supports that these were true adverse events.

Conclusions
In this study, there was a lower risk of paradoxical eczema among participants receiving IL-23 inhibitors.Factors associated with paradoxical eczema included increasing age, female sex, history of AD, and history of hay fever.These findings need replication.Future studies with more exposures and paradoxical eczema events would enable a more robust analysis of individual drugs and patient subgroups.

Table 1 .
Exposure-Level Characteristics of Included Participants, Stratified by Biologic Class Exposure

Table 3 .
Propensity Weight-Adjusted Cox Proportional Hazards Regression Survival Models for Risk of Paradoxical Eczema by Biologic Class, Biologic Drug, or Other Covariates The data for bimekizumab and tildrakizumab are not shown due to unstable effect estimates resulting from the low number of exposures and absence of paradoxical eczema events attributed to these drugs.b Single-model Cox proportional hazards regression model including all displayed demographic and clinical covariates in addition to biologic class.c The effect estimates for age represent an increase in hazard ratio per year from the base age of 18 years.
a d Includes ethnicities defined as other by the study participant or missing ethnicity (n = 20).