Factors associated with generalized pustular psoriasis progression among patients with psoriasis vulgaris in Japan: Results from a claims database study

Of those patients diagnosed with generalized pustular psoriasis (GPP) in Japan, approximately 30% have a prior psoriasis vulgaris (PsV) diagnosis. Therefore, understanding factors associated with a GPP diagnosis is essential for early diagnosis of GPP in patients with PsV. This retrospective cohort study was conducted to identify associated factors for GPP diagnosis in patients with PsV. Eligible patients with two confirmed diagnoses of PsV with/without a confirmed GPP diagnosis (International Classification of Disease 10th revision codes L40.0 and L40.1, respectively) were identified from the Japanese Medical Data Center database (JMDC) (July 1, 2005–January 31, 2019). Weighted logistic regression was used to identify associated factors (based on recorded comorbidities) between the PsV only and PsV with GPP cohorts. Odds ratios (ORs) of ≥1.5, associated with a high probability of a GPP diagnosis, were reported for factors with ≥5 patients/cohort. The time from event to GPP diagnosis was evaluated. The highest associated factor for GPP diagnosis was psoriatic arthritis (OR 20.2, 95% confidence interval [CI] 17.06–23.92, P < 0.0001), which also had the shortest time from event to GPP diagnosis (median 119 days). Other comorbidities associated with GPP diagnosis were other psoriasis, tonsillitis, and sinusitis. Treatments associated with GPP diagnosis included systemic corticosteroids (OR 2.19, 95% CI 1.98–2.43, P < 0.0001; median time from treatment initiation to GPP diagnosis 180 days). Other associated treatments (other immunosuppressants, interleukin [IL]‐17 or IL‐23 inhibitors, and phototherapy) had a delay of ≥1 year from treatment initiation to GPP diagnosis. Back pain, headache, and fever were also identified as associated with a GPP diagnosis. Patients with PsV requiring systemic therapies are more likely to receive a GPP diagnosis than those not requiring systemic treatment. These data will help identify patients with PsV at high risk of developing GPP and potentially support early GPP diagnosis.

guidelines for the management of GPP in Japan, 6 several conditions have a clinical presentation similar to that of GPP, making diagnosis challenging and leading to delays in treatment initiation. 7,8range of factors have been identified as triggers of GPP, including pregnancy, infections, and starting or withdrawing treatments such as systemic corticosteroids or anti-psoriatic therapy. 2,6,7,9,10Additionally, psoriasis vulgaris (PsV) may precede GPP, with 34%-65% of patients with GPP having prior or concomitant PsV 2,6,[9][10][11] ; it was recently shown that, in Japan, 29.7% of patients with GPP have prior or concomitant PsV. 12 PsV is a papulosquamous skin disease characterized by acanthosis, parakeratosis, and dermal inflammatory infiltrates. 2,13PsV can present with a few scattered red, scaly plaques or can involve a greater body surface area. 14Systemic symptoms include back, joint, and abdominal pain, general aching pain, headache, and fever. 15Patients may experience multiple comorbidities, including psoriatic arthritis (PsA), metabolic syndrome, cardiovascular disease, and inflammatory disorders. 15Many comorbidities and symptoms are also present in patients with GPP, 10,[16][17][18] however, it is unclear whether the presence of these symptoms or comorbidities in patients with PsV is associated with an increased risk of GPP diagnosis.
Patients with GPP have a higher clinical burden than those with PsV, 16,17 so early GPP diagnosis is essential in high-risk patients.No studies have been conducted to identify factors associated with a GPP diagnosis among patients with PsV, and the underlying mechanisms that lead to the development of GPP in this population are not fully understood.PsV is dominated by the adaptive immune response (the Th17 pathway), whereas GPP is dominated by the innate immune response (the interleukin [IL]-36 pathway), 19,20 ; therefore, it remains unknown why some patients with PsV proceed to receive a GPP diagnosis.
In this article, we present results of a retrospective study conducted using health insurance claims data from Japan to identify factors associated with a GPP diagnosis in patients with PsV.PsV with GPP cohort (whichever came first); for the PsV with GPP cohort, the date of first confirmed GPP diagnosis (Figure 1).The primary outcome was to identify factors associated with a GPP diagnosis among patients with PsV.A further exploratory outcome was the time from the factor event to a confirmed GPP diagnosis in the PsV with GPP cohort.

| Data analysis
Factors associated with a GPP diagnosis were identified using weighted logistic regression to account for the imbalance in size between the two study cohorts (PsV only: n = 28 793; PsV with GPP: n = 50).Potential factors were selected based on discussion with GPP experts in Japan and included demographics (age at index diagnosis, sex, body mass index [BMI], and pregnancy), symptoms, comorbidities, medications, and events (the number of hospitalizations and outpatient visits).Because there were many factors with an odds ratio (OR) of ≥1.0, factors with an OR of ≥1.5 (i.e., a frequency ≥1.5 times higher in patients in the PsV with GPP cohort than in those in the PsV only cohort) were defined as factors particularly strongly associated with GPP diagnosis.By contrast, an OR of ≤1.0 indicated that the odds of having a factor were lower than the odds of not having it, and therefore the factor was not included in the analysis.Significant factors associated with a GPP

| Ethics statement
This study was approved by the ethics committee of AMC Nishi-Umeda Clinic, Osaka, Japan; approval number AMC-BI-21-009.

| Study population and demographic factors associated with a GPP diagnosis
In total, 28 793 patients in the PsV only cohort and 50 patients in the PsV with GPP cohort were identified.Baseline demographics were balanced between the PsV only and PsV with GPP cohorts, with a similar mean age (44.1 and 41.3 years, respectively) and a similar proportion of male patients (62.8% and 64.0%, respectively) (Table 1).In the PsV with GPP cohort, the median follow-up (time to the second confirmed PsV diagnosis with GPP diagnosis) was 492 days (approximately 1 year and 4 months).
Age at the index date of the second confirmed PsV diagnosis and sex were tested as factors potentially associated with a GPP diagnosis; however, ORs were <1.0 (age: OR 0.966, 95% confidence interval [CI] 0.963-0.968;P < 0.0001; male sex: OR 0.887, 95% CI 0.827-0.951,P = 0.0008), therefore neither trait was considered an associated factor for GPP diagnosis.Pregnancy at any time and pregnancy within 1 year before GPP diagnosis were excluded as factors associated with a GPP diagnosis as there were fewer than five cases in the study population with each factor.BMI was excluded from the analysis as a large amount of data were missing from the database.2).Patients with these comorbidities were 1.5-20.2times more likely than those without to have a GPP diagnosis.Among these comorbidities, the time from event to GPP diagnosis was shortest for PsA, with a median of 119 days (interquartile range [IQR] 0-1421; almost 4 months); other comorbidities were first diagnosed >1 year before GPP (Table 2).Conversely, allergic rhinoconjunctivitis, asthma, and obesity were not found to be associated with a GPP diagnosis in patient with PsV (all ORs <0.75) (Table 2).GPP cohort; in order to be included in the analysis (i.e., exceed the five patient threshold), these treatments were considered as a group.b Including glucocorticosteroids.

| Treatments associated with a GPP diagnosis
time from treatment initiation to GPP diagnosis was approximately 6 months for systemic corticosteroids and just under 1 year for phototherapy, whereas the interval exceeded 1 year for both other immunosuppressants and for IL-17 and IL-23 inhibitors (Table 3).

| Events and symptoms associated with a GPP diagnosis
Clinical events such as inpatient and outpatient hospital visits can help to provide insights into the severity of disease experienced by a patient.In this study, hospitalization and outpatient visits to hospital for any reason were evaluated as potential factors associated with a GPP diagnosis, but neither had an OR of ≥1.5 (Table 4).and phototherapy), and systemic symptoms of disease (back pain, headache, and fever).Knowledge of these factors will help to identify patients with PsV who may be likely to progress to GPP and will facilitate early diagnosis and treatment of GPP, which may lead to improved clinical outcomes for these patients.These data are also important in discussions to further understand the underlying pathomechanism for developing GPP from PsV.
The association between GPP and sex has been contested in the literature.A previous study showed a predominance of male patients with GPP among those with pre-existing PsV 10 ; however, this is not the case in other studies. 8,9,16 Note: Symptoms excluded as there were fewer than five patients per cohort: chills, cough, edema, fatigue, neuropathic pain, and shortness of breath.Bold: OR ≥1.5 and P < 0.05.
entheses associated with increased mortality from cardiovascular disease, 22 indicative of severe PsV.Patients with comorbid PsA have high levels of systemic inflammation, 23 with high C-reactive protein levels.Systemic inflammation is also observed in patients with GPP and could be one reason for a high OR in patients with PsV and PsA. 6,23Previous studies have also shown that, among patients with GPP and PsA, there is a higher frequency of patients with PsV (28.4%) than without (14.3%), 10 a trend reflected in our data.This suggests that the higher rate of PsA in patients with both PsV and GPP than in those with PsV only may be because patients with PsV and PsA are more likely to develop GPP.Our finding that tonsillitis and sinusitis are also significantly associated with GPP diagnosis among patients with PsV is consistent with reports that infection triggers GPP 24  Obesity has been shown to negatively affect patients with PsV, with the pathologies of both conditions associated with a proinflammatory state. 28PsV has also been reported to be associated with allergic rhinitis and asthma. 29,30Allergic rhinitis and asthma have been classed as having strong Th2-cell-mediated responses, whereas PsV is driven by Th17 and Th1 cells. 31,32However, allergic rhinitis and asthma are heterogeneous disorders with multiple immunological phenotypes, including a Th2/Th17 reaction 29,31,[33][34][35] ; therefore, clinical linkage with psoriasis may be possible via the Th17 pathway. 29It should also be noted that allergic rhinitis is a common disease among the Japanese population, with a prevalence of approximately 40%. 36The low risk of allergic rhinoconjunctivitis, asthma, and obesity-associated patients with GPP requires further research.
It is possible that variants of psoriasis could exist on an immunological spectrum, whereby plaque psoriasis and PsV could be mediated by Th2 cells while pustular psoriasis is mediated by Th1 cells. 32Given the interplay that exists between Th1 and Th2 immune responses, it is possible that the immunopathology of GPP partially overlaps with that of PsV, but that GPP induces stronger activation which is a general limitation applicable to all database studies.There are often delays between actual GPP onset and confirmed GPP diagnosis; therefore, the factors identified in this study cannot be used This was a retrospective cohort study analyzing de-identified patient claims data from the Japanese Medical Data Center (JMDC), a commercially available employment-based insurance claims database.The JMDC compiles health insurance data for >4 million employed people in Japan.The study period was between July 1, 2005, and January 31, 2019.The study population comprised two cohorts: patients with PsV only (PsV only cohort) and patients with PsV followed by GPP (PsV with GPP cohort).The PsV only cohort included patients who had two confirmed PsV diagnoses (International Classification of Diseases 10th revision [ICD-10] code L40.0) during the study period.The PsV with GPP cohort included patients with two confirmed PsV diagnoses (ICD-10 code L40.0) and one confirmed GPP diagnosis (ICD-10 code L40.1) during the study period.Patients in both cohorts had continuous insurance enrollment after the second claim for PsV.The date of the second PsV diagnosis code was used as the index date in both cohorts.The follow-up period started at the index date and finished as follows: for the PsV only cohort, the end of continuous enrollment or maximum follow-up period in the diagnosis were those with a P value of <0.05.Descriptive analyses of the time from the first diagnosis date of the comorbidity, the first F I G U R E 1 Study design schematic.GPP, generalized pustular psoriasis; PsV, psoriasis vulgaris.treatment date of medication, the first event, and the first symptom date to GPP diagnosis were also conducted.Due to the large number of factors, the potential for multicollinearity in the model was calculated by a correlation matrix, variance inflation factor, and tolerance.Pairs of factors with a Pearson correlation value of >0.8 were considered highly correlated; factors with a variance inflation factor of >10 or tolerance of >0.1 may have had high multicollinearity with other factors.If any of the aforementioned criteria were met, the factor was excluded from the analysis after discussion with medical experts.Potential factors associated with a GPP diagnosis were also excluded if they were present in fewer than five patients per cohort.Data were analyzed using R and SAS software (version 9.4), provided by Syneos Health.

of
Th1 cells.Depending on the balance of the Th1 and Th2 pathways in individual patients with PsV, this could determine the likelihood of a subsequent GPP diagnosis.However, further investigation into the disease mechanisms of GPP and PsV is necessary to support this hypothesis.There were several limitations to this study.These included limitations of the database employed (JMDC) and the use of only a single database.First, the JMDC database holds data for 3.2% of the total population of Japan and does not include people who are unemployed, retired, working for companies who do not use the JMDC, or too unwell to work.However, the method used in this analysis can be extended, with appropriate modification, to other databases and countries.Second, claims databases do not hold granular information regarding particular characteristics of a patient's disease, such as genetic mutation status (e.g., mutations in IL36RN, CARD14, AP1S3, TNIP1, and SERPINA3 in patients with GPP), meaning that these characteristics cannot be assessed as factors associated with a diagnosis of GPP in patients with pre-existing PsV.They also do not hold clinical definitions, such as measures of disease severity and age of onset.Therefore, inferences about disease severity in this context are purely assumptive based on known variables.Another limitation was that the median time from the date of the second PsV diagnosis to the GPP diagnosis in the PsV with GPP cohort was 492 days (1.3 years), which is shorter than the median time from PsV diagnosis to GPP onset previously reported (9.5 years).10This indicates that the factors identified were based on data entry on the date of the second PsV diagnosis, rather than the date of PsV onset,
all P < 0.0001) (Table3).Only six patients received IL-17 and IL-23 inhibitors in the PsV with GPP cohort therefore, although IL-17 and IL-23 inhibitors have different mechanisms of action, these treatments were considered as a group to reach the threshold ofTA B L E 1Abbreviations: BMI, body mass index; GPP, generalized pustular psoriasis; IQR, interquartile range; N/A, not applicable; PsV, psoriasis vulgaris; SD, standard deviation.fivepatientsforinclusionintheanalysis.Fewer than five patients received apremilast and tumor necrosis factorα inhibitors in the PsV with GPP cohort, so these treatments were excluded from the analysis.Etretinate was identified as associated with a GPP diagnosis (OR 29.44, 95% CI 25.44-34.07);however,themediantimefromtreatment to GPP diagnosis was 0 days, indicating that at least half of patients received etretinate on the day of GPP diagnosis.The medianTA B L E 2Note: Comorbidities excluded as there were fewer than five patients per cohort: allergic contact dermatitis, celiac disease, Crohn's disease, chronic obstructive pulmonary disease, type 2 diabetes, diverticulitis, metabolic syndrome, myocardial infarction, osteoporosis, sepsis, stroke, ulcerative colitis, and uveitis.Bold: OR ≥1.5 and P < 0.05; bold italic: OR ≤0.75 and P < 0.05.Abbreviations: CI, confidence interval; GPP, generalized pustular psoriasis; IQR, interquartile range; OR, odds ratio; PsA, psoriatic arthritis; PsV, psoriasis vulgaris.aUnit of reference: patients with comorbidity versus those without.TA B L E 3 Treatments associated with a GPP diagnosis.PsV treatment (a Although IL-17 and IL-23 inhibitors have different mechanisms of action, only six patients (cumulative) received these treatments in the PsV with

any reason) PsV only (n = 28 793), n (%) PsV with GPP (n = 50), n (%) Total number of days by patients with PsV only, mean (SD) Total number of days by patients with PsV with GPP, mean (SD) OR (95% CI) P value
Symptoms associated with a GPP diagnosis.
The comorbidity with the strongest association with a GPP diagnosis was PsA (OR 20.2); in addition, the interval between diagnosis of PsA and GPP was short (median <4 months).PsA is a chronic, immune-mediated, inflammatory arthropathy of the joints and TA B L E 4 Events associated with a GPP diagnosis.Event (Abbreviations: CI, confidence interval; GPP, generalized pustular psoriasis; IQR, interquartile range; OR, odds ratio; PsV, psoriasis vulgaris; SD, standard deviation.TA B L E 5