Our re-analysis of the longitudinal assessment of spinal structural damage by region in a large cohort of patients with r-axSpA reveals that the retardative effect of TNFi treatment on radiographic progression (5) is not equally distributed between the spinal segments. A much greater effect can be detected in the cervical spine than the one found for the whole spine in our registry (15), explained by a smaller magnitude of the impact – not reaching statistical significance – in the lumbar spine. This result was found with progression defined as an increase in mSASSS of at least 1 or 2 units per spinal segment, as well as the formation of at least 1 syndesmophyte. It was confirmed in several sensitivity analyses: after multiple imputation of missing covariate data; after addition of disease activity as assessed by the ASDAS at treatment start; after exclusion of baseline structural damage.
The mSASSS remains the most validated and widely used method to assess spinal radiographic progression in axSpA, despite progress achieved in the area of imaging (22). The standardised clinical and radiographic assessments at regular intervals and statistical methods that take into account not only potential confounders, but also the within-patient correlation of structural damage represent important strengths of our study.
How can the finding of a comparable crude radiographic progression in the cervical and the lumbar spine over 2 years be explained in light of a more profound drug-induced inhibition of progression in the cervical spine over the same period? A higher natural progression rate in the cervical spine in patients with comparable mean symptom duration would be compatible with both findings. The fact that we found that most structural damage progression in the first 5 years of disease seems to be confined to the lumbar spine, would be reconcilable with a more important cervical progression at later time-points and with previous studies having suggested disease progression from caudal to cranial (9–11). It is important to note, that it remains unknown, whether the progression rates demonstrated in early AS studies really represent “natural” progression, given the fact that the disease-modifying effect of treatment with NSAIDs remains controversial (23, 24). This issue is discussed in more detail below, all the more we found a site-specific impact of treatment with NSAIDs comparable to the one of TNFi.
Several, mutually not exclusive hypotheses can be put forward to explain a differential inhibition of progression in the cervical versus the lumbar spine. The first hypothesis is related to the fact that structural changes seem to start in the lumbar spine. Inhibition of progression might not be possible any more if certain reparative changes have already been initiated, and this might occur at an earlier time-point in the lumbar spine. Magnetic resonance studies have demonstrated that syndesmophyte formation is more likely to occur at VCs in which inflammation has been replaced by fatty degeneration, than at VCs with persistent inflammation (25). The fact that structural damage seems to start in the caudal spine would imply, that in the first few years after start of symptoms, inhibition of progression would only be detectable in the lumbar spine. Given the long mean symptom duration in our cohort, the number of patients with early disease was too low to allow testing this assumption. In line with this argumentation, inhibition of progression should be detected at all spinal levels if TNFi are initiated early on and sustained remission is achieved. Indeed, almost no progression could be detected in patients reaching an ASDAS < 1.3 before a next radiographic interval at both cervical and lumbar level in our study. The regional difference in inhibition of progression was most clearly depicted in patients with persisting high disease activity despite bDMARD treatment. A second hypothesis involves the presence of degenerative spinal disease that might interfere with the assessment of axSpA-induced osteoproliferative changes. Indeed, degenerative changes overlap with axSpA-associated lesions even in early disease and most frequently involve the more distal aspects of the spine (13, 14). However, it has been shown that trained readers are able to distinguish between axSpA-associated versus degenerative lesions (14, 18). Our primary readers were involved in these studies, rendering this hypothesis less probable, though not absolutely impossible. Thirdly, mechanical strain was shown to be able to enhance new bone formation (12). Biomechanical forces are larger at the level of the lumbar spine and might lead to enhanced progression in the caudal region of the spine and potentially counteract pharmacological inhibitory effects. However, there were no clues for progression being more important in the lumbar spine during late disease. Moreover, the number of physical exercise sessions per week, introduced as a proxy for physical strain in our investigation, did not significantly impact on the results. Finally, site-specific developmental differences might be involved. Joint-specific anatomical diversity has been demonstrated for synovial fibroblasts and for cartilage with regard to the expression of homeobox (HOX) family genes (26, 27), which are crucial for the embryonic development of limb and vertebrae. Imprinted developmental differences could therefore also control site-specific activation of axSpA-relevant pathways in the entheses along the different regions of the spine.
Irrespective of the mechanism(s) leading to differential regional inhibition of progression as delineated above, the applicability of our findings is manifold. The clinical relevance of retardation of radiographic progression in axSpA upon treatment with TNFi has been questioned (5). Indeed, progression is not linear in the individual patient. Moreover, it only affects a proportion of patients in a given interval of 2 years. Our data suggesting that inhibition of progression might be restricted to certain spinal segments further challenges its clinical relevance. However, the inhibitory effect on cervical structural damage is of such magnitude (68% odds reduction of radiographic progression for TNFi), that it seems of utmost clinical relevance for the patients concerned, particularly for rapidly progressing patients. The more important potential to detect inhibition of spinal radiographic in the cervical spine might also have an influence on the demonstration of the disease-modifying effect of other drug classes. The issue of inhibition of radiographic progression by nonsteroidal anti-rheumatic drugs (NSAIDs) remains, as already mentioned, controversial. Two randomized controlled trials investigating the effects of on-demand use versus continuous use of different classes of NSAIDs on spinal progression in AS reached opposite conclusions (12, 13). In our main models, treatment with NSAIDs reached a retardative impact on radiographic progression in the cervical spine that was comparable in size to treatment with TNFi at this spinal segment, while no significant effect of NSAIDs was detected in our original analysis of total mSASSS. The fact that we could not calculate a NSAIDs intake score as recommended by the ASAS (28) and no information was available on the type of NSAID used (traditional NSAIDs or coxibs) represent major limitations of our current analyses. Time-varying treatment with NSAIDs was only available as “yes/no” at start of each interval and this information was included as such in our models, assuming that rheumatologists mainly considered a “yes” when the patient used NSAIDs on a regular basis and that this was continued in the following radiographic interval. Whether the putative symptom duration-dependent differential inhibitory effect on progression in the cervical versus the lumbar spine also applies to treatment with NSAIDs and might explain the previously contrary results obtained in trials assessing the impact of NSAIDs on progression, therefore, remains unclear. Site-specific re-analysis of previously performed trials, as well as newly designed trials seem warranted to confirm our findings. Our findings could also be of importance when comparing the disease-modifying capacities of different classes of biologic or targeted-synthetic disease modifying drugs in axSpA in head-to-head clinical trials. Site-specific assessment might improve detection of potential differences in progression, particularly as the expected differences might be rather small (28). As demonstration of causality is difficult to achieve in an observational context (5), only comparative head-to-head trials will be able to provide a definitive answer to the conundrum whether disease modification through inhibition of osteoproliferation is possible in axSpA, provided that a difference in progression between patients treated with different drugs can indeed be detected.