What is behind this scientifically fallible approach to invent a diagnosis of ‘physician-reported spondylitis’?

It has been discussed for a long time whether axial PsA is a separate entity of PsA or if it should just be seen as a form of axSpA. There were some arguments favouring the latter view,8 9 in line with a recent study stressing the influence of HLA-B27 on inflammatory axial disease in PsA.10 Differences between axial PsA and axSpA have also been highlighted11 even though the radiographic part of the New York criteria has been found useful also for axial PsA.12 Altogether, this means that regarding the existential question of one versus two diseases, the jury is still out.

In addition, the unexpected failure of the anti-IL-23 RCTs in axSpA1 2 has undoubtedly been a disappointment for the pharmaceutical companies involved, and the suggestion of improvements in a few typical axSpA measures may have incentivised the post-hoc analyses of the PsA trials. There is no trial on guselkumab in axSpA though. However, the STAR study with guselkumab in axial PsA is currently recruiting patients (NCT04929210).

What is the content of these two post-hoc analyses ?

Taken together, we have two formal RCTs, which have clearly shown that ustekinumab works in PsA, and we have two post-hoc analyses of these RCTs, looking at patients with back pain present on request, as well as axSpA instruments suggesting increased disease activity: as if these patients with PsA were patients with axSpA. The main difference between the two studies is that the second one4 focused on 127 anti-TNF-naive patients of PSUMMIT 1 and PSUMMIT 2, while the first one3 had looked at all patients with ‘signs of axial involvement’ as explained (n=256). In this post-hoc study, a modification of a (published) modified version of the BASDAI (mBASDAI) was used, which simply excluded BASDAI question 3 (the original modification excluded questions 3 and 413).

Another disease activity measure for axSpA, the disease activity score ASDAS has been shown to be superior to the BASDAI but both measures are still widely used (boxes 1 and 2). The ASDAS was used in the second post-hoc analysis.4 Both, BASDAI and ASDAS had been evaluated in patients with AS, implying that these patients had been included on the basis of back pain, stiffness and decreased mobility. However, the performance of the BASDAI when evaluated in patients with PsA was not convincing.14

The baseline demographics3 reveal a rather typical peripheral PsA population: 45 years old on average, 45% women, less than 7 years PsA duration and on average 14 swollen and 24 tender joints, half of them with dactylitis. More than 80% of patients had signs of enthesitis (mean Maastricht Ankylosing Spondylitis Enthesitis Score (MASES) around 5), and not <90% had confirmed peripheral erosions. The mean Psoriasis Area and Severity Index (PASI) was 14.5 and half of the patients was on treatment with methotrexate. In spite of this typical peripheral PsA-Gestalt, the BASDAI (only the back pain question: how would you describe the overall level of AS neck, back or hip pain you have had during last week?) was more than 6/10, which made the authors suggest that axial involvement was present.3

The second study4 showed slightly lower values; 20% of patients were on glucocorticoids. The mean ASDAS was 3.9 and the BASDAI as well as the mBASDAI were approximately 7, while only a minority (25%) was human leukocyte antigen (HLA)-B27+. Looking at efficacy, there were no differences between the ustekinumab doses, and the mean change of BASDAI was −1.8 at week 12 and −2.1 at week 24, independent of HLA-B27 status.4 Since international recommendations prescribe a change of at least two for an individual patient with axSpA to prove efficacy, this appears to be a rather mild improvement. However, this is only approximative since these are changes at the group and not at an individual level.

Of the 1120 patients in the two DISCOVER studies, 312 (28%) were included in this analysis,6 of whom 118 were on placebo, and 194 in the guselkumab every 4 weeks or 8 weeks groups with 61% being male, mean age 45 years. Out of 190 patients with known HLA-B27 status 30% were positive. There was one major methodological difference regarding patient selection in this post-hoc analysis, since patients were included if there was ‘evidence’ of axial disease documented by previous imaging or pelvic radiography at screening consistent with sacroiliitis, as confirmed by the investigator6—which may or may not be any better than simply claiming ‘physician-reported spondylitis’. In any case it is known that diagnostic decisions only based on imaging of the sacroiliac joints are problematic, and that is even more true for PsA where degenerative changes may cause problems for the differential diagnosis. This may also have influenced the results of a study which reported, based on sacroiliac radiographs, underdiagnosed axial disease especially in female patients with PsA.15

In this study6 increased disease activity—as if these patients were active patients with axSpA—was assumed if the (m)BASDAI was >4 as in the ustekinumab studies described above.3 4 At week 24, mean changes from baseline in BASDAI and ASDAS were significant: −2.7 and −1.4 in both guselkumab groups versus −1.3 and −0.7 in the placebo group, respectively. The results for mBASDAI and spinal pain were similar. Improvements with guselkumab were seen at week 24 independent of HLA-B27 status.6 The latter is rather an argument against the claim to have successfully treated axial PsA.10

Is there an effect of anti-IL-23 on supposed but unproven axial inflammation in PsA ?

On the basis of a rather negative evaluation, the absence of evidence that BASDAI works in PsA16 17 and the fact that none of the patients in PSUMMIT 1 and PSUMMIT 2 had been included because of back pain but rather because of peripheral arthritis, known to be frequently associated with fatigue, generalised entheseal pain and morning stiffness, we find it difficult to believe that the mild change in the mBASDAI observed in a post-hoc selection of patients with PsA, can be attributed to assumed but unproven inflammation in the axial skeleton of patients with the leading symptom of peripheral PsA. We rather think that the improvement of back pain is a bystander effect of general improvement in patients who had severe peripheral PsA, with involvement of many joints and skin and a high burden of disease. It is well known that pain-spreading mechanisms (central sensitisation) may cause back pain which may improve on improvement in general well-being.

In conclusion, we think that post-hoc analyses of peripheral PsA trials do not suffice to prove that ustekinumab and guselkumab (or anti-IL-23 treatment in general) are efficacious for ‘real inflammation’ in the axial skeleton of patients with PsA. Far better definitions for axial PsA are needed to be used for inclusion in clinical trials and we strongly support the ongoing Assessment of Spondyloarthritis International Society (ASAS)/Group for Research and Assessment of Psoriasis and Psoriatic Arthritis (GRAPPA) initiative in this regard.

Disease-specific measures, such as BASDAI and ASDAS, should be applied in the context in which they have been developed, in this case proven axial inflammation. Once axial inflammation of patients with peripheral PsA has been proven, for example, by MRI, BASDAI and ASDAS may be appropriate tools to follow these patients over time, but not before that has been achieved. We also do not apply DAS28 in patients with fibromyalgia without a diagnosis of rheumatoid arthritis.

In conclusion, we propose to abandon the term ‘physician-reported spondylitis’. The likelihood that axial PsA is much different from axSpA in terms of treatment response in patients with proven axial inflammation is, in our opinion, rather low.