Impact of Season and Other Factors on Initiation, Discontinuation, and Switching of Systemic Drug Therapy in Patients with Psoriasis: A Retrospective Study

This study investigated whether systemic drug prescribing for psoriasis varies by season and other exacerbating factors. Eligible patients with psoriasis were assessed for each season for initiation, discontinuation, and switching of systemic drugs. A total of 360,787 patients were at risk of initiating any systemic drugs in 2016‒2019; 39,572 patients and 35,388 patients were at risk of drug discontinuation or switching to a biologic and a nonbiologic systemic drug, respectively. The initiation of biologic therapy in 2016‒2019 peaked in spring (1.28%), followed by summer (1.11%), fall (1.08%), and winter (1.01%). Nonbiologic systemic drugs followed a similar pattern. Those aged 30‒39 years, male, those with psoriatic arthritis, those who live in the South region, those who live in areas with lower altitudes, and those who live in areas with lower humidity had higher initiation with the same seasonality pattern. Discontinuation of biologic drugs peaked in summer, and switching of biologics was highest in spring. Season is associated with initiation, discontinuation, and switching, although seasonality pattern is less clear for nonbiologic systemic drugs. Approximately 14,280 more patients with psoriasis in the United States are estimated to initiate a biologic in spring than in other seasons, and over 840 more biologic users switched in spring than in winter. The findings may provide evidence for healthcare resource planning in psoriasis management.


INTRODUCTION
Psoriasis is a chronic systemic inflammatory disease affecting over 7.5 million adults and 0.9 million children in the United States (US) (Armstrong et al., 2021;Paller et al., 2018). The common symptoms of psoriasis include scaling of the skin, itching, and erythema (Dubertret et al., 2006;Feldman et al., 2017;Pariser et al., 2016). Important comorbidities are cardiometabolic disease, psoriatic arthritis (PsA), gastrointestinal disease, kidney disease, malignancies, infections, and mood disorders (Egeberg, 2016;Takeshita et al., 2017). Psoriasis is associated with reduced work productivity and QOL, which are more affected by increasing psoriasis severity (Villacorta et al., 2020;Korman et al., 2016Korman et al., , 2015. The annual US cost of psoriasis amounted to approximately $112 billion in 2013 (Brezinski et al., 2015). Those with comorbidities are associated with a higher economic burden (Feldman et al., 2017).
Several factors have been shown to contribute to psoriasis flare. These include obesity/weight gain (Aune et al., 2018;Wolk et al., 2009), tobacco smoking (Aune et al., 2018;Wolk et al., 2009), infections (Skudutyte-Rysstad et al., 2014, stress (Chandran and Raychaudhuri, 2010;Farber and Nall, 1984), and low humidity (Chandran and Raychaudhuri, 2010;Farber and Nall, 1984). Few studies have examined seasonality in psoriasis flare (Harvell and Selig, 2016;Kardes ‚ , 2019;Wu et al., 2020); however, there are seasonal fluctuations in these known triggers of body weight (Fahey et al., 2020) and cigarette consumption (Dubertret et al., 2006;Momperousse et al., 2007;Rachakonda et al., 2014). Regional differences have been found in healthcare resource use for psoriasis (Nguyen et al., 2020), psoriasis severity, comorbidities, and treatment response in the US (Enos et al., 2021), and increasing latitude is associated with increased prevalence (Gutierrez et al., 2017;Springate et al., 2017). Over 17% of patients with psoriasis are treated with systemic therapy or phototherapy, an indicator of higher levels of disease severity (Crown et al., 2004). The objectives of this study were to explore seasonal patterns in the initiation, discontinuation, and switching from systemic drugs among patients with psoriasis and to assess whether the change in systemic drugs varies by patient characteristics and other potentially exacerbating factors. The study findings may provide real-world evidence for the use of systemic drugs for psoriasis and help with psoriasis management and healthcare resource planning.

Patient population
Patient flow chart for each cohort is presented in Figure 1. A total of 360,787 patients with psoriasis were at risk of initiating any systemic drugs in 2016-2019 (see Table 1). They were on average aged 54.2 years (SD ¼ 18.2), 52.5% were female, and 6.4% had PsA at cohort entry. The South region of the US had the most patients (44.9%), followed by the Midwest (22.3%), the West (19.1%), and the Northeast regions (13.4%). Two thirds came from latitude 39.85 N; over 60% were from states with average annual relative humidity less than 77.1%. A total of 39,572 patients were at risk of discontinuing or switching from any biologics (see Table 1). They consisted of patients who initiated biologics or continued their existing biologics. They were younger (mean age 48.9 vs. 56.6 years) and more likely to be male (52.8 vs. 43.4%) than the 35,388 patients who were at risk of discontinuing or switching from any nonbiologic systemic drugs. Among biologic users, the ratio of patients who were at risk of discontinuing or switching from TNF-a inhibitors, IL-12/23 inhibitors, IL-17 inhibitors, and IL-23 inhibitors was roughly 7:3:2:1, highest for those who entered the market earliest, lowest for the newest drug class. Figure 2 shows the impact of season on the initiation of systemic drugs in patients with psoriasis in 2016-2019. Initiation of biologics is on average higher than that of nonbiologic systemic drugs (0.9-1.4% vs. 0.8-1.1%). The initiation of nonbiologic systemic drugs peaked in spring (0.9-1.1%). The initiation of any biologic followed a pattern similar to that of nonbiologic systemic drugs, with the incidence highest in spring (ranging from 1.1 to 1.4%) and lower in other seasons (ranging from 1.0 to 1.3%). Table 2 presents the incidence and 95% confidence interval (CI) of initiation of biologic therapy and stratified by different factors in patients with psoriasis in 2016-2019. The mean incidence of initiation of biologic therapy (95% CI) in 2016-2019 was highest in spring (1.28% [1.25-1.30%]), followed by that in summer (1.11% [1.08-1.13%]), fall (1.08% [1.06-1.10%]), and winter (1.01% [0.99-1.03%]). Figure 3 shows that biologics initiation by drug class also peaked in spring. Among biologics, the incidence of initiation seemed to be highest for TNF-a inhibitors, followed by that in IL-12/IL-23 inhibitors, IL-17 inhibitors, and IL-23 inhibitors. To account for the clustering within the data due to repeated measures in Table 2, the meta-regression analysis results in Table 3 confirms that the incidence of initiation of biologics was the highest in spring among the four seasons. The incidence of initiating biologics overall was 13% lower in summer (relative rate [RR] ¼ 0.87, 95% CI ¼ 0.82e0.92), 16% lower in fall (RR ¼ 0.84, 95% CI ¼ 0.84e0.95), and 21% lower in winter (RR ¼ 0.79, 95% CI ¼ 0.69e0.91) than in spring. The initiation trend was consistent by biologic class and patient characteristics (sex, age, and, PsA status).

Impact of season on initiation of systemic drugs
Figures 4-9 show that when stratified by sex; age; diagnosis of PsA; and US region, latitude, and humidity, the peak in the initiation of any biologic therapy was spring, followed by summer and fall or winter. The mean incidence of initiation of systemic drugs was higher in males than in females in all seasons (1.16-1.45% in males vs. 0.88-1.12% in females), with seasonal differences quite similar in males and females. When stratified by age, the incidence of initiation of systemic drugs was highest in those aged 30-39 years, followed by those 40-49 years, 20-29 years, 50-59 years, 60-69 years, <20 years, and 70 years. The seasonal pattern is very clear in all age subgroups except in those aged <20 years, who had higher initiation in summer and fall. Patients with PsA are at least five times more likely to initiate any biologics than those without (mean incidence ¼ 4.19-5.10% in PsA vs. 0.78-0.99% in non-PsA). The seasonal pattern is more obvious in patients with PsA than in those without, with seasonal differences ranging from 0.3 to 1.2% in PsA and from 0.1 to 0.2% in those without PsA. The incidence of any biologic initiation appeared to be higher in patients in the South and Midwest regions than in those in the West and Northeast regions (mean incidence ¼ 1.16-1.52% in the South vs. 1.10-1.30% in the Midwest vs. 0.76-1.00% in the West vs. 0.71-0.89% in the Northeast). Initiation of biologics appeared to be higher in latitude <39.85 N than in latitude 39.85 N (1.0-1.5% vs. 0.8-1.4%) and in lower (<77.1%) versus higher (77.1%) humidity regions.     Figure 10 shows the impact of season on discontinuation of systemic drugs in patients with psoriasis in 2016-2019. The incidence of discontinuation was lower for biologics (ranging from 11.3 to 15.0%) than for nonbiologic systemic drugs (ranging from 14.9 to 19.2%) across seasons, with the incidence difference between nonbiologic systemic drugs and biologics, largest in spring. The peak of discontinuation was inconsistent for biologics or nonbiologic systemic drugs; discontinuation peaked in two winters for nonbiologic systemic drugs and in two summers for biologics. Further analysis of biologic discontinuation by drug class based on a four-season mean showed that discontinuation for TNF-a inhibitors and IL-17 inhibitors appeared to be highest in winter and that discontinuation for IL-12/IL-23 inhibitors appeared to be highest in summer (see Table 4). The mean incidence of discontinuation of biologics seemed to be highest in summer overall and in stratified analyses, except for patients aged 60 years and patients with PsA, who had the highest discontinuation rate in the winter (see Table 4). The incidence of discontinuation of biologic drugs does not seem to differ by sex but seems lower in patients with PsA, patients in the West and Midwest regions, patients in latitude 39.85 N, and patients in higher humidity regions. In addition, the incidence of discontinuation of biologics among adults aged 20 years decreases with increasing age, lowest in those who were aged 70 years (see Table 4). For nonbiologic systemic therapy, both the mean incidence of discontinuation and stratified analyses seemed to be higher in spring and summer (data not shown). In Table 5, the results from meta-regression analysis show that the incidence of discontinuation of biologics overall was 7% higher in summer than in spring (RR ¼ 1.07, 95% CI ¼ 1.07e1.07) but was not significantly different in fall (RR ¼ 0.99, 95% CI ¼ 0.93e1.06) and winter (RR ¼ 1.00, 95% CI ¼ 0.90e1.12). The discontinuation trend was consistent by biologic class and patient characteristics (sex, age, and PsA status).
Impact of season on switching of systemic drugs Figure 11 shows that the incidence of switching was higher for biologics (ranging from 0.03 to 0.15%) than for nonbiologic systemic drugs (ranging from 0 to 0.10%) across all seasons. The switching of nonbiologic drug therapy seemed to be on lack of clear pattern, with a tendency of being lower

H Liang et al.
Seasonality of Systemic Drug Therapy in Psoriasis in winter and higher in spring and summer. The switching of biologics appeared to be lower in winter and higher in spring. Stratified analyses by season did not show clear seasonality, and the incidence of switching was not different by any stratifying variables, except that switching appeared to be higher in PsA than in non-PsA (data not shown). The mean incidence of switching from biologic therapy (95% CI) in 2016-2019 in winter was 0.07% (95% CI ¼ 0.04-0.09%) as opposed to 0.12% (95% CI ¼ 0.09-0.15%) in spring and 0.08% (95% CI ¼ 0.06-0.11%) in summer and fall (see Table 6). Stratified analysis clearly showed that the mean incidence of switching from biologics appeared highest in spring, except for those aged <20 and 70 years. The incidence of switching from TNF-a inhibitors and IL-17 inhibitors appeared to be higher than that of switching from IL-12/IL-23 inhibitors. The results of the meta-regression analysis in Table 7 does not indicate a statistically significant seasonal effect on the incidence of switching for biologics, although the point estimate indicated a trend of higher switching in summer. The highest incidence of switching in spring (see Table 6) supports the peak in the initiation of biologic drugs in spring.

DISCUSSION
This study found that the initiation of systemic drugs for psoriasis peaked in spring and then declined in summer and fall. This pattern was consistent for all biologics and nonbiologic systemic drugs and within strata of potentially exacerbating factors for psoriasis. The incidence of initiation of any biologic was much higher in males than in females, highest in those aged 30-39 years than in other age groups; those with comorbid PsA were more likely to receive systemic drugs. The incidence of any biologic initiation appeared to be higher in the South and Midwest regions, low latitude regions, and low humidity regions than in other regions. In addition, this study found that discontinuation of systemic drugs peaked in winter or summer and that switching from systemic drugs tended to be lower in winter and higher in spring and summer.
No previous studies were identified that evaluated seasonality and environmental factors in relation to initiation and discontinuation of systemic therapies for psoriasis. However, several studies have reported a seasonal relationship to psoriasis in general (Hancox et al., 2004;Harvell and Selig, 2016;Jensen et al., 2022;Kardes ‚ , 2019;Wu et al., 2020). Two studies reported that the frequency of Google search data for the term psoriasis and related terms peaked in the late winter/early spring and troughed in the late summer/ early fall (Kardes ‚ , 2019;Wu et al., 2020). Our study found that the initiation of systemic drugs peaked in spring, which was supported by another study (Hancox et al., 2004). Hancox et al. (2004) reported that there is seasonal utilization of dermatologic care in the US and that dermatologic office visits peaked in spring and troughed in fall (33.8 vs. 20.3% of annual visits).
Several studies identified a seasonal association with the worsening of psoriasis symptoms. One study from India reported that 42% of patients with psoriasis worsened in winter versus 8% in summer, whereas 43% improved in the summer, and only 7% improved in winter (Kaur et al., 1997). An online survey of adults with psoriasis from 15 countries reported that 77% of respondents reported seasonal variation of psoriasis exacerbation most notably in winter (67.1%) compared with 23.8% in summer, 7% in spring, and 2.1% in autumn (Ferguson et al., 2021). A nationwide survey of over 12,000 patients with psoriasis in China found that season change was the most frequently reported cause of relapse or aggravation (60.2%). Nearly half of reports about the weather as an aggravating factor were related to the winter season (48.8%), followed by spring (23.1%), autumn (17.1%), and summer (8.4%) (Chen et al., 2017). A retrospective study of 2,270 patients with psoriasis in China found that a total of 53.2% reported the seasonal pattern of disease, with psoriasis exacerbation in fall/winter (Zheng et al., 2021). Oral or biologic treatment may be initiated if psoriasis is too extensive for topical therapy or refractory to topical therapy and phototherapy (Menter et al., 2009  for patients with psoriasis (Chen et al., 2017). Our study found the highest initiation rate in the South region. South region in the US has been associated with the highest proportion of patients with obesity and very severe psoriasis (body surface area > 20%) in the US Corona Psoriasis Registry (Enos et al., 2021), which reported that psoriasis severity and comorbidities differed among US geographic regions (Enos et al., 2021). Our finding of seasonal variation in the initiation of systemic drugs for psoriasis has not been reported in the literature and has important implications. A recent systematic review of 13 studies reported that about 50% of patients with psoriasis were stable and showed no seasonal differences between seasons and that approximately 30% improved in summer and 20% performed better in winter (Jensen et al., 2022). Guidelines on the management of psoriasis are suggested to add seasonality so that treatment and patient education may be considered to prevent disease worsening. Although the relative seasonal change for the initiation of systemic drugs appeared small in this study (0.1-0.4% for biologics and 0.1-0.3% for nonbiologic systemic drugs), the mean reduction for biologic initiation from 1.28% in spring to 1.11% in summer could lead to an estimated decrease in the number of biologic initiators for over 14,280 patients, considering that there are over 7.5 million adults and 0.9 million children with psoriasis in the US (Armstrong et al., 2021;Paller et al., 2018). Therefore, although the relative seasonal change is small, the absolute seasonal change has clinical significance because the findings provide evidence for healthcare resource planning in psoriasis management. In addition, we found that the mean incidence of biologic switching appeared to be highest in spring. This change in systemic drug use may not reflect the disease severity because patients' behaviors around wanting to be free of psoriasis and be able to wear short sleeves/swim in warmer months may impact their use of systemic drugs. However, the initiation of systemic drugs in the study was determined on the basis of the date of systemic drug dispense or administration. It is typically related to a clinical visit or a medical claim with a psoriasis diagnosis. The timing of initiation of a new systemic drug may indicate a potential psoriasis flare for the patient. Whether a higher proportion of patients discontinued biologics owing to psoriasis improvement in the summer is unknown. Thus, the findings from this study may be used for hypothesis generating, and further patient-levelbased research is needed.
Our finding that the incidence of the initiation of biologic drug peaks in spring is supported by the higher incidence of discontinuation and switching from nonbiologic systemic drugs in spring and the higher incidence of switching from biologics in spring. Although the 95% CIs overlapped, the mean increase for biologic switching from 0.07% in winter to 0.12% in spring would result in an estimated increase in the number of biologic switchers for over 840 patients, considering a US psoriasis population of approximately 8.4 million (Armstrong et al., 2021;Paller et al., 2018) and that about 20% are moderate or severe (Menter et al., 2008). We found that the mean incidence of biologics discontinuation appeared to be highest in summer, except for those aged 60 years and those with PsA. Switching or discontinuation from biologics could be due to primary ineffectiveness, secondary loss of response, side effects, patient preferences, comorbidities, and economic burden (Bayaraa and Imafuku, 2019;Florek et al., 2018). Treatment switch is frequent in psoriasis, with 50% of traditional systemic-treated patients switching to a biologic; the age of these patients tended to be younger. Conversely, 25% of the biologic group transitioned to traditional oral systemic therapy; these patients tended to be older and have a longer duration of disease (Tabolli et al., 2015).
This study found that the initiation of any biologics varied by patient characteristics. Patients' preferences for certain treatments may depend on age, sex, comorbidities, disease duration, and prior treatments (Florek et al., 2018). Abbreviations: CI, confidence interval; IL-17i, IL-17 inhibitor; IL-23i, IL-23 inhibitor; TNFai, TNF-a inhibitor; NA, not applicable; IL-12/23i, IL-12/ 23 inhibitor. Note: Robust variance estimation was used to estimate the covariance matrix of the correlated coefficients in the meta-regression accounting for the clustering within the data owing to repeated measures (Fisher et al., 2017). The meta-regression model was fitted with the incidence transformed into a natural logarithmic scale and season as a fixed effect. The higher incidence of initiation of any biologics in males than in females may reflect patient preference and that men with psoriasis are more likely to have severe psoriasis than women (Florek et al., 2018;Hägg et al., 2013). In addition, men with psoriasis have been found to be more concerned about efficacy than women (Kromer et al., 2015). The finding that the highest incidence of initiation of systemic drugs occurred in patients aged 30-39 years is also supported by the literature (Bayaraa and Imafuku, 2019;Florek et al., 2018;Geale et al., 2016;Gorelick et al., 2019). Psoriasis impacts the patients' self-esteem and QOL (Kubanov et al., 2018;Nazik et al., 2017;Pariser et al., 2016). At least 90% of young patients with psoriasis value clear skin, sustained response, and rapid onset of action (Gorelick et al., 2019). As people age, patients with psoriasis have fewer opportunities to access biologic medications (Geale et al., 2016).
This study has several strengths. First, the exposure season is clearly defined. This study assessed treatment patterns using data from all the four years, and similar patterns were observed. In addition, the seasonal pattern in the initiation of systemic drugs is supported by the discontinuation and switching data. Second, effect modifiers were stratified to assess the association between seasons and treatment patterns of systemic drugs, and similar patterns held after stratification. Third, study findings may not only inform healthcare providers and patients in their decision making in patient disease management but also generate, to our knowledge, previously unreported research ideas.
A few limitations need to be mentioned. First, although a clear association has been identified between seasons and initiation, discontinuation, and switching of systemic drugs in patients with psoriasis, no causal relationship can be made. Second, patients with rheumatoid arthritis, ankylosing spondylitis, inflammatory bowel disease, and other conditions for which certain biologics may be indicated were not excluded from the study. Therefore, there might be some misclassification in the initiation, discontinuation, and switching of certain biologics. However, the impact could be minimal given that the prevalence of these conditions is less common. Third, the association between season and outcomes was not  assessed in a mixed-effect regression model, where the season can be a fixed effect, and a patient can be a random effect to derive the seasonal percentages while accounting for the correlation among repeated measures. Future studies may be conducted to assess whether the findings from this study are true in patient-level drug utilization studies in psoriasis population (which will allow patients to be followed up without interruption of season end) and whether season is an independent factor to predict change in systemic therapy while controlling for other factors. Finally, patients in this database are covered by commercial insurance. Healthcare insurance coverage may affect treatment patterns as well (Armstrong et al., 2017). The study findings may not be generalizable to those without healthcare insurance or covered by different healthcare insurance. In summary, a seasonal pattern of initiation of systemic drug therapies for psoriasis is identified in this study. However, this finding needs to be interpreted with caution because initiation of a systemic therapy does not indicate a reactive response to a loss of disease control and increased symptoms. Discontinuing ineffective drugs and switching to alternative systemic drugs among individuals who use medications in reaction to psoriasis symptoms might be a key component in reducing the risk of psoriasis worsening. Future research may focus on specific systemic drug survival rates and switching patterns among patients with certain comorbidities.

Study design
The study is a retrospective ecological study of individuals with psoriasis identified in the Optum Clinformatics Data Mart. Two cohorts were generated for each season of 2016-2019: patients with psoriasis who were eligible for initiating a systemic drug and those with psoriasis who were prevalent users of a systemic drug. The outcomes were the incidence of initiation of a systemic drug

Study setting
The Optum Clinformatics Data Mart is commercially available in the US. It contains longitudinal commercial and Medicare Advantage health plan data from 50 US states since 2000. The claims data include member eligibility, medical and pharmacy claims, and inpatient confinements. It covered approximately 16 million annual lives associated with United Healthcare plans for over 80 million unique lives. On the basis of the Health Insurance Portability and Accountability Act in the US, all personal identifiers for patients are anonymized by applying specific algorithms. For example, patient names, social security numbers, and five-digit zip codes were removed from the databases. The date of birth was converted to the year of birth before data release. Thus, institutional review board review was unnecessary.

Study population
The study population included patients with psoriasis, defined as those with a diagnosis of psoriasis (     referred to person-time in 2016-2019 that a patient contributed after a psoriasis diagnosis and before the systemic drug initiation. Patients who had a psoriasis diagnosis and continuously used a systemic agent were eligible for discontinuing or switching from the systemic drug. Only person-time that met these criteria was considered at risk of discontinuation or switching. In evaluating initiation, discontinuation, and switching of systemic drugs, a 30-day gap was allowed between prescription refills. If the same drug has multiple prescriptions with overlapping periods, the treatment duration was the sum of those periods. An at-risk patient was followed until the earliest occurrence of initiating, switching from, or discontinuing a systemic agent; death; disenrollment; end of the study; and ceased contributing person-time. Numerators cohorts came from the denominator cohorts and were patients who initiated, switched from, or discontinued a systemic drug during the season of interest and were also assessed separately for the season of interest. The numerator cohorts were determined using treatment data at least 6 months before the season and treatment data during each season in 2016-2019. Figure 12 presents how denominators and numerators for initiation, discontinuation, and switching are estimated in each season.

Outcomes
The outcomes were incidence of initiating, switching from, and discontinuing a systemic drug, first assessed for each drug of interest and then grouped by drug classes. Specifically, initiation of a systemic drug was defined as the first date of that drug dispensing or administration in the season under study. To determine initiation, the use of systemic medications was assessed using data 180 days before CED. Switching from one systemic drug to another was defined as the first date for a prescription stream for an alternative drug that occurred within 30 days before a patient exhausted or discontinued his initial therapy. Discontinuation of a systemic drug was defined as the last day of a therapy stream, which required at least 30 days free of any biologic or nonbiologic therapy after the end of the therapy stream. For all these outcomes, the first event for drug initiation, discontinuation, and switching was assessed in each season, ranging from 90 to 92 days. It is less likely that a clinician would change a patient's systemic therapy two times in such a short time window. Torres et al. (2021) reported that at 12 months, IL-23 inhibitors had over 90% cumulative probability of drug survival, whereas an IL-17 inhibitor had the lowest cumulative probability of 85.5% for drug survival. However, patients who received the second or third biologic and so on for 180 days in the following seasons would qualify to be assessed for further initiation, switching, or discontinuation.
The systemic drug classes assessed comprised TNF-a inhibitors, IL-12 and IL-23 inhibitors, IL-17 inhibitors, IL-23 inhibitors, any biologics, and nonbiologic systemic immunosuppressants. Table 8 presents the generic names for systemic medications and their use frequencies and prescription length. The number of days covered with each biologic was captured on the basis of its mode of administration. Self-administered biologics dispensed at the pharmacy were identified from prescription claims using National Drug Codes, and days of supply were used to calculate the number of days covered by each prescription. Biologics that required infusion under the supervision of medical professionals were identified from medical claims using the Healthcare Common Procedure Coding System codes. The administration date and the assigned days' supply for each administration based on recommended dosage regimen were used to calculate the estimated medication end date. For biologics that were found in both prescriptions claims and medical claims, the medication end date was estimated using the prescription fill date and days of supply or administration date and assigned days of supply. Abbreviations: CI, confidence interval; IL-17i, IL-17 inhibitor; IL-23i, IL-23 inhibitor; TNFai, TNF-a inhibitor; NA, not applicable; IL-12/23i, IL-12/ 23 inhibitor. Note: Robust variance estimation was used to estimate the covariance matrix of the correlated coefficients in the meta-regression accounting for the clustering within the data due to repeated measures (Fisher et al., 2017). The meta-regression model was fitted with the incidence transformed into a natural logarithmic scale and season as a fixed effect.

Data analysis
A patient flow chart was generated for each study cohort. Patient demographics were summarized with descriptive statistics. The incidence of initiating a systemic drug was calculated using the number of patients who initiated a systemic drug during the season divided by the number of patients with psoriasis who were eligible for initiating a systemic drug on CED. Similarly, the incidence of switching from or discontinuing a systemic drug was calculated using the number of patients who switched from or discontinued a systemic drug during the season divided by the number of patients with psoriasis who used the systemic drug before CED. The incidence of initiation, discontinuation, and switching and 95% CIs were calculated separately for all seasons in 2016-2019. Of note, winter 2016 was from January 1, 2016 to February 29, 2016; thus, the incidence for winter 2016 needs to be interpreted with caution. In addition, the mean incidence and 95% CI for winter through the fall season in 2016-2019 were also calculated. To determine a seasonal trend, we compared the 95% CIs for the incidence of initiation, discontinuation, and switching for each season. If the 95% CIs among seasons do not overlap, then the two incidences are considered statistically different. To account for the clustering within the data due to repeated measures, metaregression analyses with robust variance estimation were conducted for biologics. The robust variance estimation considers the covariances between outcomes from different studies and provides an estimation of covariance matrix. The estimates from each season are considered separate studies for initiation, discontinuation, and switching. Therefore, data from 16 studies (seasons) were used in each meta-regression analysis. Meta-regression models were fitted with all the incidences transformed to natural logarithmic scale and season as a fixed effect for the incidence of initiation, discontinuation, and switching for all biologics and stratified by patient characteristics, region, latitude, and humidity. Metaregression analysis results were presented in Tables 3, 5, and 7. In addition, line graphics were used to show the seasonal trend for systemic drugs overall and the seasonal trend stratified by each covariate. Finally, to show clinical significance, the change in the number of biologic initiators from spring to summer in 1 year was estimated on the basis of the absolute difference in the mean incidence of biologic initiation between spring and summer in 2016-2019 and estimated psoriasis population in the US (Armstrong et al., 2021;Paller et al., 2018); the change in the number of biologic switchers from winter to spring in 1 year was estimated on the basis of the absolute difference in the mean incidence of biologic switching between winter and spring in 2016-2019 and estimated moderate-to-severe psoriasis population in the US (Armstrong et al., 2021;Menter et al., 2008;Paller et al., 2018).
To assess the impact of season on the initiation, switching, and discontinuation of systemic drugs, analyses were performed in the AETION Evident Platform, a cloud-based scientifically validated software that transforms real-world data into transparent, reliable, and replicable real-world evidence (Folkerts et al., 2020;Garry et al., 2019;Seesaghur et al., 2021). All analyses were stratified by possible effect modifiers of the association between season and initiation, discontinuation, and switching of systemic drugs. No patientlevel datasets can be downloaded from the seasonality analysis module of the AETION Evident Platform so mixed-effect regression analysis of the patient-level data is not feasible. However, Rstudio, version 3.6.0, was used for the entry of data from 16 seasons for biologic initiation, discontinuation, and switching. Robumeta package in R was used for meta-regression analyses (Fisher et al., 2017). Abbreviations: CI, confidence interval; IL-17i, IL-17 inhibitor; IL-23i, IL-23 inhibitor; TNFai, TNF-a inhibitor; NA, not applicable; IL-12/23i, IL-12/ 23 inhibitor. Note: Robust variance estimation was used to estimate the covariance matrix of the correlated coefficients in the meta-regression accounting for the clustering within the data due to repeated measures (Fisher et al., 2017). The meta-regression model was fitted with the incidence transformed into a natural logarithmic scale and season as a fixed effect. For IL-17 inhibitors, only the data from 2018 and 2019 are used, to minimize the effect of multiple new products launched in 2015 and 2016.

Data availability statement
No large datasets were generated or analyzed during this study. Minimal datasets necessary to interpret and/or replicate data in this paper are available on request to the corresponding author.