The identification of SPMS transition remains a crucial clinical challenge since the absence of a widely recognized definition in differentiating relapsing and progressive forms. Nowadays, neurologists diagnose SPMS retroactively considering a history of progressive impairment following an initial relapsing phase. [17, 18] The difficulty of detecting SPMS patients also seems to affect whether they will be characterized as active or nonrelapsing [8].

In addition, using adapted HERCULES criteria, in this multicenter retrospective study we identified a distinct nrSPMS phenotype characterized by older age and fewer relapses after conversion. Only a small proportion of post-conversion relapses resulted in RAW, supporting the concept that disability accumulation in this group is largely driven by PIRA mechanisms. These findings are consistent with prior studies reporting a predominance of nonactive SPMS and highlighting PIRA as the main driver of disability. A recent study [8] identified SPMS patterns in the cohort of the Danish MS Register, outlining a higher percentage of nonrelapsing SPMS patients at the time of conversion and a higher cumulative probability of being nonactive by the end of follow-up. The different criteria applied to the RISM cohort could explain the discrepancy with this study. The population of nonactive SPMS in the RISM database has been also recently described [19]: compared to active SPMS, the patients in this cohort were older and had a longer disease duration.

Considering the results of the study of Chisari et al. [19], the proportion of patients presenting PIRA at 24 months follow-up was comparable in the active and nonactive groups, confirming that PIRA is largely responsible for progression [2, 4, 20, 22].

A recent study by Portaccio et al. [22] highlighted that a single PIRA event, particularly in the case of sustained PIRA, is sufficient to start progression in most of the patients. In addition, in our cohort, PIRA events were more frequent than RAW and occurred earlier than formal SPMS conversion, reinforcing the notion that PIRA may precede and signal the transition to secondary progression. Time to first PIRA was shorter than time to SPMS diagnosis, and accounting for PIRA reduced the association between different SPMS definitions, suggesting that early PIRA substantially influences disease course classification. These observations support the view that the occurrence of PIRA should prompt closer monitoring for progression and potential treatment reassessment.

Diagnostic performances of standardized definitions for the diagnosis of SPMS are the subject of compelling and recent debate. A recent study analyzed data from five European MS registries and showed different proportions of identified SPMS patients employing three objective classification methods, the MSBase algorithm [23], the EXPAND criteria, and a decision tree-based algorithm [24]. The sensitivity differed between the methods (47.4% EXPAND, 75.7% MSBase algorithm, 83.5% decision tree), while the specificity resulted similar. The overall accuracy of each method compared to the clinical SPMS assignment was 79.1% for EXPAND, 85.4% for the MSBase algorithm, and 83.7% for the decision tree classifier [24]. Braune et al. outlined that the specificity of the MSBase criteria for SPMS was also high (89.6% in the group of patients diagnosed with SPMS and 96.2% including RRMS patients without transition), in parallel with low values of sensitivity (32%) and accuracy (61.4%) [25]. The HERCULES criteria, compared to the ND, applied to our cohort resulted in low sensitivity (37.6%) and high levels of specificity (93.4%), in line with previous results. Nonetheless, it is important to emphasize a high level of accuracy of 91.1% of HERCULES-adapted criteria. Considered together, these performances indicate that the HERCULES-adapted criteria offer a highly specific and therefore useful approach to identify well-defined SPMS cases with high confidence. The low sensitivity should be viewed in the context of the reference standard, as clinician-defined SPMS is often heterogeneous and delayed. These measures were not intended to support individual-level risk prediction, but to quantify overall agreement and classification performance in the absence of a true gold standard for SPMS diagnosis.

The observed differences between ND and HERCULES-defined SPMS populations should be interpreted considering the intrinsic differences between the two diagnostic frameworks. By design, these definitions capture inherently different patient populations. This fundamental distinction could explain several of the observed findings, including younger age and higher inflammatory activity in ND-defined SPMS. Conversely, the lower frequency of PIRA observed in ND-defined SPMS may reflect reduced exposure time for relapse-independent progression, as disability changes occurring in temporal proximity to relapses cannot be classified as PIRA. In this context, the higher number of SPMS diagnoses and the earlier timing of SPMS identification observed with HERCULES criteria are particularly noteworthy because these findings suggest that criteria-based definitions may reclassify early PIRA-related disability accumulation as SPMS in the contemporary treatment era. This concept aligns with the ongoing debate surrounding the concept of how specific PIRA truly is for inflammation-independent progression. Emerging evidence supports a shift from purely clinical definitions of PIRA toward a biologically driven framework grounded in MS pathogenic mechanisms and MRI correlates, moving beyond traditional clinical descriptors [30].

We also performed an analysis of geographical discrepancies in SPMS definition. Patients followed in MS Centres in Northern Italy show a lower percentage of PIRA and HERCULES defined patients, compared to MS Centres of Central and Southern Italy. A previous work of the RISM assessed geographical socio-economic factors and characteristics of local MS centers in Italy, underscoring their role of primary factor influencing phenotype at initial neurological evaluation [26]. Further research will be useful to explore possible determinants and consequences of regional variability.

Some key limitations of the study must be acknowledged. A primary limitation is that the algorithm depends solely on EDSS records, with all the related limits of this scale. [27] Groups differed in disability severity, with EDSS < 3.0 in the ND reference group and EDSS 3.0–6.5 in the HERCULES definition. Observed discrimination may partly reflect disability level rather than disease course, potentially leading to an overestimation of the ability to identify SPMS. Furthermore, as discussed, the stringent criteria of the data-driven definition adopted in our analysis may explain the different proportion of nrSPMS patients reported in other studies. [8] Potential cohort-related biases should be considered when interpreting these findings. Specifically, the lower SPMS diagnostic rates observed in our cohort compared with classic natural history studies, in which approximately 23–25% of patients converted to SPMS over 10–11 years, likely reflect differences in follow-up, clinical characteristics and study design [31, 32].

Consistent with other retrospective studies based on disease registries, data incompleteness and entry inaccuracies cannot be excluded. A major limitation of the adapted HERCULES definition is the exclusion of radiological activity, as MRI data are not standardized across the RISM network, resulting in a classification based solely on clinical parameters. This distinction is clinically relevant and should be acknowledged when interpreting our findings. Despite using the ND as the reference definition for comparison, we cannot rule out the possibility that the neurologists treating patients included in our cohort underdiagnosed SPMS because of a diagnostic delay that may have been caused by the hesitancy to diagnose secondary progression [28] and the later onset of the progressive phase [29]. Furthermore, in this study we did not account for DMT class or treatment patterns in our cohort and relapse suppression from high-efficacy therapies may have influenced nrSPMS classification, representing a potential confounder to consider when interpreting these results. No formal sample size calculation was performed, as this was a retrospective, exploratory, registry-based study.

Our findings underline the importance of using algorithms for SPMS diagnosis, particularly when they are applied as standardized end points in epidemiological research based on MS registry data. In addition, the study is based on the RISM patient population, which is highly representative of the Italian MS patient population. [16] With participants from 125 MS centers (about 81% of all centers of RISM network), the population of RISM included in this study is also highly representative, being geographically widely distributed.

In conclusion, applying nrSPMS criteria adapted from the HERCULES trial to our cohort reliably identified a distinct nrSPMS subgroup, accounting for 7.89% of our study population. NrSPMS patients were less likely to experience relapses after SP conversion and had even fewer RAW events, while PIRA emerged as the predominant CDA event, being markedly more frequent than RAW and occurring earlier in the HERCULES group than in the ND. Overall, these findings reinforce the need to prevent disability accumulation primarily driven by smoldering neuroinflammation.