2.1 Study design and setting

This retrospective cohort study compared the short and long-term skeletal outcomes of adult patients with psoriasis who initiated RZB versus MTX. Data were extracted from TriNetX, a global federated health-research network that aggregates de-identified electronic health records from more than 100 health-care organizations. Because only anonymized, aggregate data were used, institutional review board approval was not required.

2.2 Comparator rationale

We selected methotrexate (MTX) as the active comparator because it is a commonly used conventional systemic therapy for psoriasis and is frequently considered prior to biologic escalation. This study was designed to address a specific clinical question: among adults with psoriasis initiating systemic therapy, whether initiation of IL-23 inhibition (risankizumab) versus MTX is associated with differences in subsequent fracture and arthroplasty outcomes. Comparisons with other biologic classes (for example, IL-17 inhibitors) were outside the scope of the present analysis and are noted as an important direction for future head-to-head real-world studies.

2.3 Participants

Adults (≥18 years) with a documented diagnosis of psoriasis (ICD-10-CM L40.0) who received a first prescription or administration of RZB or MTX between January 1 2014 and June 14 2025 were identified. The date of first exposure was defined as the index date. First prescription or administration reflects the first recorded exposure within the TriNetX network and does not guarantee treatment naïvety outside participating health systems. This is addressed in the limitation. Patients were required to have at least 2 years of follow-up and were observed for a maximum of 5 years. Individuals were excluded if, before the index date, they had a record of any major fracture (shoulder/upper arm, lumbar spine or pelvis, femur, or forearm) or arthroplasty of the hip or knee. Diagnostic and procedural codes used for inclusion and exclusion criteria are provided in Supplementary Table 1. Within the eligible cohort, treatment groups were defined on the basis of drug exposure codes: RZB (ATC: L04AC23; RxNorm: 2068445) and MTX (ATC: L01BA01; RxNorm: 6657).

2.4 Demographics and propensity-matched cohort

Demographics included age, sex, race, ethnicity, and body-mass-index category. Baseline clinical covariates included comorbidities known to be relevant to fracture risk, osteoporosis and osteopenia diagnoses, and other comorbidities captured by ICD-10-CM codes (Supplementary Table 1). Comorbidities were defined by diagnosis codes recorded on or before index and were not contingent on receipt of a specific treatment. TriNetX's built-in logistic-regression algorithm was employed to generate a 1:1 propensity-matched sample of the selected demographics using nearest-neighbor matching with a caliper width of 0.01. Covariate balance was confirmed when absolute standardized differences were <0.10.

2.5 Outcomes

Outcomes were grouped into short-term (90-day) events to evaluate early safety and acute events after treatment initiation, and into long-term (2- and 5-year) clinical and disease-related complications to evaluate intermediate and long-term follow-up within available data. 90-day complications included myocardial infarction, pulmonary embolism, deep-vein thrombosis, acute liver failure, stroke, skin infection, psoriatic arthritis, joint infection, sepsis, emergency-department visit, and inpatient hospitalization. 2-and 5-year outcomes included skin infection, psoriatic arthritis, joint infection, sepsis, total hip arthroplasty, total knee arthroplasty, revision arthroplasty, periprosthetic joint infection, mechanical loosening, and other device-related complications. Outcomes were identified via validated ICD-10-CM diagnosis codes and Current Procedural Terminology (CPT) procedure codes (Supplementary Table 1). Fractures were identified using ICD-10-CM diagnosis codes at prespecified anatomic sites. Because EHR coding does not reliably capture the mechanism of injury across sites, fracture outcomes reflect all coded fractures at these sites rather than confirmed low-energy fragility fractures; this limitation is discussed.

2.6 Statistical analysis

All analyses were conducted within the TriNetX analytics platform. Cohorts were propensity score-matched 1:1 using nearest-neighbor matching with a caliper of 0.01 on the propensity score. Covariate balance was assessed using standardized mean differences, with values < 0.10 indicating acceptable balance. Risk ratios with 95% confidence intervals were calculated for each outcome at each prespecified timepoint. Categorical variables were compared using chi-square tests and continuous variables using Student's t-tests, with P < 0.05 considered statistically significant.

2.7 Study type

This manuscript follows the required structure for human observational research submissions and aligns with the journal's expectations for study design transparency, including adherence to the STROBE guidance for cohort studies.

3.1 Demographic and clinical characteristics that distinguish adult psoriasis patients who initiate MTX from those who initiate RZB

In the prematched analysis, the RZB cohort consisted of 5451 adults with psoriasis, whereas 54,402 patients received MTX. RZB patients were slightly younger (50.4 ± 15.2 years vs. 51.8 ± 17.0 years, p < 0.001) and were more often male (47.7% vs. 37.9%, p < 0.001). They also had higher prevalences of hypertension (27.1% vs 23.8%), chronic kidney disease (4.4 % vs 2.8 %) and obesity (BMI >30 kg m⁻² = 23.3% vs 26.5%), but lower rates of diabetes and heart failure (all p < 0.001, Table 1). After 1:1 propensity score matching, 5448 patients remained in each cohort. Matching eliminated most baseline imbalances: age (50.4 ± 15.2 vs 50.1 ± 15.7 years, p = 0.233), sex (women, 50.9% vs 50.3%, p = 0.527), and the distribution of race or ethnicity was comparable. Residual differences were small; heart failure (3.0% vs 2.2%, p = 0.007) and overweight or obesity (14.8% vs 13.2%, p = 0.019) remained slightly more common among RZB recipients, while type 2 diabetes was modestly lower (14.2% vs 12.1%, p = 0.001). All other demographic, comorbidity, and BMI categories showed standardized differences below 0.1, confirming adequate balance (Table 1).

Short term (90-day) and long term (2 and 5-year) effects of each therapy on clinical, infectious, and skeletal outcomes, including fracture risk and need for major joint arthroplasty.

3.2 90-Day clinical complications

At 90 days, clinical outcomes were mostly comparable between RZB and MTX groups. Notably, the incidence of new-onset psoriatic arthritis was significantly lower in the RZB cohort (1.0% vs. 5.7%; RR 0.18, 95% CI 0.128–0.252; p < 0.001). Other outcomes, including myocardial infarction, pulmonary embolism, deep vein thrombosis, acute liver failure, stroke, skin infection, sepsis, emergency department visits, and inpatient hospitalizations showed no significant differences (Table 2). Short term psoriatic arthritis and infectious outcomes at 90 days are also shown in Fig. 1.

Close

Fig. 1 Psoriatic arthritis and infectious outcomes over time in patients treated with risankizumab or methotrexate. Bar graphs show the percentage of patients with (A) psoriatic arthritis, (B) skin infection, (C) joint infection, and (D) sepsis at 90 days, 2 years, and 5 years. Asterisks indicate comparisons with p < 0.05. RZB, risankizumab; MTX, methotrexate.

Export to PPT

3.3 2-Year clinical outcomes

2-year outcomes demonstrated continued significant differences in psoriatic arthritis occurrence, favoring the RZB group (5.9% vs. 16.7%; RR 0.351, 95% CI 0.302–0.407; p < 0.001). Additionally, the incidence of sepsis was lower in the RZB cohort (1.2% vs. 1.7%; RR 0.709, 95% CI 0.516–0.974; p = 0.033). Other outcomes, including skin infection, joint infection, shoulder/upper-arm fracture, femoral fracture, lumbar/pelvis fracture, hip arthroplasty, knee arthroplasty, and periprosthetic joint infections, did not show significant differences (Table 3). These 2 year patterns are included in Figs. 1 and 2.

Close

Fig. 2 Fracture and hip arthroplasty outcomes over time in patients treated with risankizumab or methotrexate. Bar graphs show the percentage of patients with (A) shoulder or upper arm fracture, (B) lumbar or pelvis fracture, (C) femoral fracture, and (D) hip arthroplasty at 2 years and 5 years. Asterisks indicate comparisons with p < 0.05. RZB, risankizumab; MTX, methotrexate.

Export to PPT

3.4 5-Year clinical outcomes

At 5 years, RZB treatment continued to show significantly lower rates of psoriatic arthritis (9.5% vs. 22.6%; RR 0.418, 95% CI 0.371–0.471; p < 0.001), skin infection (9.0% vs. 12.1%; RR 0.748, 95% CI 0.662–0.844; p < 0.001), joint infection (0.2% vs. 0.7%; RR 0.33, 95% CI 0.172–0.633; p < 0.001), and sepsis (2.3% vs. 3.3%; RR 0.696, 95% CI 0.553–0.875; p = 0.002) (Table 4). RZB was also associated with lower 5-year rates of shoulder/upper-arm fracture (0.7% vs. 1.5%; RR 0.491, 95% CI 0.335–0.718; p < 0.001), lumbar/pelvis fracture (0.8% vs. 1.4%; RR 0.539, 95% CI 0.370–0.787; p = 0.001), and hip arthroplasty (0.9% vs. 1.4%; RR 0.631, 95% CI 0.437–0.910; p = 0.013). Rates of femoral fracture, knee arthroplasty, and periprosthetic joint infections were similar between groups (Table 4). Temporal trends in psoriatic arthritis, infectious complications, and fracture or arthroplasty events from 90 days through 5 years are illustrated in Fig. 1 (Fig. 1A–D) and Fig. 2 (Fig. 2A–D).

4.1 Demographic and clinical characteristics that distinguish adult psoriasis patients who initiate MTX from those who initiate RZB

Our analysis of patient demographics revealed notable differences in the pre-matched cohorts. Patients initiating RZB were typically younger, comprised a higher percentage of males, and were more often non-Hispanic compared to those on MTX. This demographic distribution suggests potential prescribing patterns where younger, male, and non-Hispanic patients may be preferentially offered newer, more specialized treatments like RZB. This observation warrants further investigation to understand if there are implicit biases or socioeconomic factors influencing access to advanced therapies, potentially limiting their prescription to women or certain ethnic groups. Furthermore, the finding that patients on RZB, even after matching for key covariates, tended to have a higher comorbidity burden (e.g., higher prevalences of hypertension, chronic kidney disease, and overweight/obesity) indicates that physicians might be reserving more specialized or alternative treatments for individuals presenting with more complex health profiles and greater disease severity. This aligns with a clinical strategy where sicker patients, who may not respond adequately to conventional therapies, are transitioned to or initiated on biologics that may ultimately influence downstream orthopedic risk profiles.⁹

4.2 Short-term (90-day) and long-term (2- and 5-year) effects of each therapy on skeletal health, including fracture risk and need for major joint arthroplasty

A key finding of this study is the sustained protection offered by RZB against the development of PsA. From 90 days out to 5 years, the incidence of new-onset and overall PsA was significantly lower in the RZB cohort compared to the MTX group. This reinforces the known efficacy of IL-23 inhibitors in managing the arthritic component of PsO.¹⁰ Although the incidence of PsA during follow-up was defined as the first new PsA diagnosis occurring after the index date, baseline PsA may be present in a subset of patients. Thus, interpreting PsA-related outcomes may warrant some caution, as they may reflect either incident disease or disease flares.

Regarding bone health, our study found that while RZB offered benefits in terms of significantly decreased risk of shoulder/upper-arm and lumbar/pelvis fractures, and a decreased risk of hip arthroplasty at 5 years, these benefits were generally not drastically different from those observed with MTX for other major fracture types (e.g., femoral fracture) and knee arthroplasty. Rare methotrexate-associated osteopathy with stress fractures has been described and could contribute to observed differences in specific fracture patterns, although it cannot be reliably identified in aggregated EHR data. This suggests that although RZB provides a protective effect on bone health at selected skeletal sites, it may not offer a substantial added benefit at other regions compared with MTX, one of the most conventional and widely used disease-modifying antirheumatic drugs (DMARDs). Importantly, taking RZB did not appear to increase the risk of overall progression to poor joint and bone health in patients, which is a reassuring finding for a relatively newer therapeutic agent.

4.3 Limitations

This study is subject to inherent limitations associated with its retrospective, observational design. Although propensity matching balanced measured baseline characteristics, residual confounding may persist, particularly if patients initiating risankizumab achieved better inflammatory control than patients initiating methotrexate. Others include lifestyle factors (e.g., physical activity levels, dietary calcium intake, smoking status) and treatment adherence, which cannot be entirely excluded. Thus, these associations may require cautious interpretation because treatment selection in routine care is strongly influenced by factors that also affect fracture risk, including disease severity, systemic inflammatory burden, functional status, and concomitant medication use. A key limitation is potential misclassification and overlap between PsO and PsA at baseline. If baseline PsA or axial involvement differed between cohorts, this could influence both fracture risk and arthroplasty outcomes, independent of treatment effect. Glucocorticoid exposure, particularly systemic therapy, is an important determinant of fracture risk and may not be fully captured or timed relative to outcomes in de-identified EHR data. Additionally, the relatively recent FDA approval of RZB (approximately 5 years from study onset) meant that reliable long-term data for periods exceeding 5 years were not available within the TriNetX network. This limits our ability to assess the very long-term impact of RZB on skeletal outcomes and major joint procedures, highlighting a need for future studies with extended follow-up durations.