Structural damage as an outcome measure in clinical trials
The utility of radiographic assessment of structural damage as an outcome in PsA is illustrated by several phase 3 clinical trials of biologic agents used to treat PsA (Table
2) [
15,
17,
19‐
32,
34,
36]. The first studies of the biologic agents etanercept [
15] and adalimumab [
17] in PsA used a modified Sharp score for PsA to assess the effect of therapy on structural damage. In the study evaluating etanercept vs placebo, the primary radiographic endpoint was the annualized rate of change in the modified Sharp score for PsA at 6 and 12 months [
15]. At 12 months, etanercept led to inhibition of radiographic progression in the hands and wrists compared with worsening in the placebo group (
P = .0001); annualized changes in the erosion score and joint space narrowing scores were also significantly different in the two groups. In the ADEPT study of adalimumab vs placebo, inhibition of structural damage, as measured by the Sharp method for PsA at week 24, was a primary efficacy endpoint [
17]. At week 24, active treatment demonstrated significant inhibition of structural damage progression compared with placebo (
P < .001). Significant differences were also observed in erosion and joint space narrowing scores.
Since then, other trials of biologics in PsA have used the SvdH scoring system (Table
2). The IMPACT studies were some of the first studies to use the SvdH method to assess structural damage in patients with PsA being treated with a biologic agent (infliximab) [
19,
20]. The primary outcome was mean progression of structural damage as demonstrated by the mean changes from baseline in the total SvdH score, with positive changes from baseline indicating progression of structural damage. The studies showed that the SvdH scoring method was appropriate for use in PsA; however, the usefulness of scoring features characteristic of PsA (e.g., hand DIP joints, pencil-in-cup changes, gross osteolysis) was limited, as had been observed previously when using the Sharp method in PsA [
15,
17]. This was mainly due to the low progression rate of these features over 6 to 12 months. Importantly, the IMPACT studies showed that biologic treatments were able to inhibit radiographic progression as early as 6 months, and therefore assessment of structural damage in clinical trials can be performed after 6 months as opposed to waiting 1 to 2 years to assess damage. These studies also suggested that for ethical reasons, patients in clinical trials should not be offered placebo for long periods of time but should be allowed to receive active treatment after a 6-month period.
In line with these findings, the GO-REVEAL [
21] and GO-VIBRANT [
22] studies of golimumab showed that 24 weeks of follow-up are sufficient for radiographic damage as an outcome. In these studies, the change from baseline in the SvdH scoring method for PsA of the hands and feet at week 24 was one of the two coprimary endpoints (Table
2). Scores for PsA-specific radiologic damage (e.g., DIPs, pencil-in-cup, and gross osteolysis deformities) were also included. As seen in the IMPACT studies, patients who originally received placebo and later crossed over to receive active treatment had more structural damage after 1 year than did patients who originally received active treatment, suggesting a benefit associated with earlier treatment.
As new targeted therapies for PsA have been developed, inhibiting radiographic progression has become essential for demonstrating efficacy and disease-modifying activity. For instance, the phase 3 studies PSUMMIT-1 and PSUMMIT-2 of the anti–interleukin (IL) 12/IL-23 antibody ustekinumab assessed changes from baseline in radiographic progression at week 24 using the SvdH method for PsA (Table
2) [
25]. Findings from this study showed that inhibiting targets other than tumor necrosis factor (i.e., IL-12 and IL-23) could also lead to improvements in PsA and inhibition of radiographic progression. In the FUTURE studies of secukinumab (an anti–IL-17 inhibitor), inhibition of radiographic progression was a key secondary endpoint [
26,
27,
34]. Patients treated with secukinumab had significantly less radiographic progression, defined as the change from baseline in SvdH score at week 24. Inhibition of radiographic progression was sustained up to 2 years for both erosion and joint space narrowing [
26,
37]. These studies demonstrated that targeting IL-17A was another therapeutic option for patients with PsA. Similar assessments of radiographic progression were conducted in the SPIRIT-P1 study of ixekizumab, another IL-17 inhibitor [
28,
32], and in studies of the Janus (JAK) inhibitor tofacitinib [
30] and the T cell modulator abatacept [
29] (Table
2).
The impact of methotrexate on radiographic outcomes has also been assessed. The phase 3 randomized study SEAM-PsA, which examined the efficacy of methotrexate and etanercept monotherapies vs methotrexate in combination with etanercept [
31,
36] (Table
2), found that concomitant methotrexate did not lead to significant changes in radiographic outcomes, consistent with previous observations [
15,
26]. However, in studies of golimumab [
21] and infliximab [
20], treatment group differences were larger in patients receiving concomitant methotrexate.
Limitations and considerations in the analysis of radiographic progression in clinical trials
Given that it is based on subjective interpretation of radiographic changes, the assessment of radiographic progression is subject to possible measurement error and variability. In general, the presence of radiographic damage in clinical trials is preferably assessed by 2 or 3 central readers to ensure reliable information [
38]. Additionally, a mean change of ≤ 0.5 in total score (vs 0) is preferred when the mean of two readers is used to determine the absence of radiographic progression.
A main challenge in assessing radiographic progression is detecting a treatment effect [
39,
40]. Ethical considerations in clinical trials limit the duration of placebo treatment, impacting the ability to detect radiographic progression in this group. Additionally, patients randomized to placebo are allowed to switch to active treatment due to lack of efficacy or to discontinue the study before radiographic progression is assessed or discernable. These low rates of radiographic progression in the control group and incomplete radiographic data may impact the statistical power of these trials to detect a treatment effect [
39].
Another important challenge is mitigating the impact of missing radiographic data. Linear extrapolation and interpolation is a widely used approach and requires radiographic data from ≥ 2 time points [
23,
27]. Linear mixed-effects models, which account for important cofactors such as previous treatment and baseline values, are another approach that can maximize the statistical power of clinical trials and mitigate the effect of missing data. These methodologies have been used in various studies. For example, in the IMPACT-2 study, missing data were imputed using linear extrapolation or the median of the change in total scores based on all patients within the same methotrexate stratification (i.e., a median of 0) [
20]. In the GO-REVEAL study, linear extrapolation was used, and if data were insufficient for linear extrapolation, the median change in total SvdH score was used to replace missing data [
21]. However, imputation of median scores is no longer applied. In the FUTURE5 study, radiographic data were analyzed by a linear mixed-effects model that excluded data after escape for patients treated with placebo who received escape therapy at week 16. The model assumed approximately linear progression over time and estimated a difference in rates of progression over 24 weeks to compare treatment groups [
41]. However, it is important to note that imputation methodologies may significantly influence interpretation of radiographic outcomes, as seen in the RAPID-PsA phase 3 study of certolizumab pegol, where (incorrect) imputation methodologies resulted in a high degree of progression in all arms [
23]. In this study, not imputing missing data for patients with ≤ 1 radiograph or 2 radiographs < 8 weeks apart and linear extrapolation in patients with two radiographs ≥ 8 weeks apart were found to be the most appropriate methods for the primary analysis.
In addition to imputation methodologies, enriching for patients who are at high risk of radiographic progression may increase the power of a study to detect treatment effects [
39]. This could be achieved by increasing the number of patients who present with predictive factors for radiographic progression or by indirectly enriching the data through post hoc analyses. Although predictive factors for radiographic progression in PsA are limited, systemic inflammation as indicated by elevated baseline C-reactive protein has been shown to be a strong independent predictor of radiographic progression and may serve as a way of enriching for high-risk patients [
42,
43]. Similarly, the existence of radiographic damage is another predictive factor. Patients with damage are more prone to develop more damage, especially in the presence of an elevated C-reactive protein.
Another limitation of clinical trials that assess structural damage in PsA is that they tend to focus on radiographic progression in peripheral joints only. For instance, structural changes associated with enthesitis, including anabolic bone formation, are not generally assessed, and no systematic method of measurement is currently available [
12]. Similarly, progression in the axial skeleton is not commonly measured. Different scoring methods for assessment of axial involvement are available [
44]; however, they have not been used in large clinical trials of PsA so far. A further limitation is that minimum clinically important differences for radiographic scores have not yet been established.
Future areas of research
Scoring methods for other imaging techniques (i.e., micro-computed tomography scan, ultrasound, MRI) should be developed, validated, and further tested for their propensity to assess structural damage in PsA [
11]. So far, only MRI has been used and is scored using the Psoriatic Arthritis Magnetic Resonance Image score (PsAMRIS) [
45,
46]. PsAMRIS is a semiquantitative scoring system that has been developed for PsA by the international MRI in arthritis group of OMERACT (Outcome Measures in Rheumatology) [
45,
46] and is the most validated system available for the evaluation of inflammatory and structural changes in the hands of patients with PsA [
45]. PsAMRIS has been used in a few small trials in which a significant improvement in inflammatory parameters was demonstrated following treatment with a biologic agent; however, bone damage parameters, such as bone proliferation and erosion, showed little change over time [
45]. New developments in MRI approaches, such as dynamic contrast-enhanced MRI and digital automated analysis, may improve MRI techniques [
47].
Additionally, further refinement of radiographic scoring methods specific for PsA merits investigation. For example, radiographic scoring methods could be improved by accounting for the progression of new bone growth, which is characteristic of PsA. Another area of development could be the use of artificial intelligence to assess structural damage on radiographs.