Discussion

In this systematic review and meta-analysis, we synthesized the available evidence on the prevalence and diagnostic performance of susceptibility-based MRI biomarkers—PRLs and the CVS—in POMS. Both biomarkers were frequently detectable in POMS, although substantial variability was observed across studies for CVS-related outcomes. Although only a few studies included POMS-mimicking conditions, available evidence suggests that PRLs may have high specificity for distinguishing POMS from other disorders, whereas CVS, despite being more prevalent in POMS, appears less specific.

PRLs in POMS

Across seven studies including 149 patients with POMS, approximately 72% harbored ≥1 PRL. Between-study heterogeneity was low, and pooled estimates were robust across sensitivity and leave-one-out analyses, indicating relatively consistent findings despite differences in MRI acquisition protocols and analytical approaches. PRL assessment showed generally good inter- and intra-rater reliability in the studies reporting these metrics, supporting the technical feasibility and reproducibility of PRL identification in pediatric populations.

In AOMS, the presence of ≥1 PRL has recently been estimated at 52% (95% CI 47–58) in a meta-analysis of 37 studies including 3180 patients, although with substantial between-study heterogeneity [26]. Notably, the pooled prevalence observed in POMS appears higher, even though some pediatric studies relied on routine clinical MRI protocols acquired at 1.5-T. While no significant effect of age on PRL prevalence was reported in the aforementioned AOMS meta-analysis [26], some studies suggest that PRL burden may decrease with increasing age and disease duration, consistent with the dynamic nature of PRLs and potential rim fading over time [27].

Beyond prevalence, two studies showed that PRLs were absent in MOGAD, NMOSD, and ADEM, resulting in 100% specificity for POMS, while sensitivity ranged from 69 to 92%. In AOMS, PRL specificity has been estimated at approximately 98% when excluding Susac’s syndrome, whereas reported sensitivity varies widely across studies (from 10 to 100%) [27], with the largest study to date reporting a prevalence of 52% [28]. The pediatric findings align with the high specificity observed in adults, although they should be interpreted cautiously given small sample sizes and limited representation of MS mimics. Interestingly, a recent study of 65 children with MOGAD identified no PRLs, providing additional support for the high specificity of PRLs in POMS [29]. Further support for a distinct biology between POMS and pediatric MOGAD is a recent study showing the absence of slowly expanding lesions (SELs), another marker of chronic active lesions, in pediatric MOGAD, while they were detected in POMS [30].

CVS in POMS

In contrast to PRLs, CVS appeared to be less specific to POMS. The pooled mean proportion of CVS-positive lesions per POMS patient was approximately 58%, but between-study heterogeneity was very high, and pooled estimates varied considerably depending on analytical choices. Similarly, the proportion of patients fulfilling the 40%-CVS rule showed wide variability and unstable pooled estimates.

It is important to note that while CVS is detected in MOGAD and other conditions, the ability to compare the relative prevalence of CVS between different diagnoses is challenging methodologically. CVS detectability is highly sensitive to MRI field strength, spatial resolution, susceptibility contrast, and post-processing techniques. The included studies used heterogeneous MRI protocols, spanning 1.5-T and 3-T systems and a variety of susceptibility-based sequences. In addition, lesion detection criteria and analytical strategies differed substantially across studies, and inter-rater reliability was high in some studies but poor in one, [17] underscoring the operator-dependent nature of CVS assessment and the need for dedicated training in CVS interpretation.

In AOMS, the pooled proportion of CVS-positive lesions has been estimated at 73% (95% CI 67–79) in a meta-analysis of 29 studies, with reported values ranging from 58% at 1.5-T, to 74% at 3-T, and 82% at 7-T [31]. Notably, this meta-analysis also reported substantial between-study heterogeneity, mirroring the variability observed in the POMS population [31]. A study directly comparing POMS and disease-duration-matched AOMS reported a non-significantly higher proportion of CVS-positive lesions in AOMS and a significantly higher proportion of AOMS patients fulfilling the 40%-CVS rule [16]. Lower sensitivity of CVS in POMS may reflect the presence of large confluent lesions – particularly in periventricular regions – which are considered by CVS methodological criteria to be ineligible for CVS scoring.

Only one study evaluated the Select-6 rule in POMS, and only one assessed the ≥50% CVS-positive lesion threshold recommended for pediatric populations. In the latter, sensitivity decreased markedly, with only 23% of patients fulfilling the criterion, suggesting that this cutoff may be overly restrictive in some pediatric cohorts. At present, the available evidence is insufficient to determine whether specific CVS thresholds should be preferentially applied according to lesion burden or other clinical characteristics in POMS.

Only two studies formally assessed the diagnostic performance of CVS in POMS, both reporting a higher CVS prevalence in POMS than in MS-mimicking conditions. However, a recent study in a large pediatric MOGAD cohort found that 35.5% and 22.5% of patients fulfilled the 40% and 50% CVS-positive lesion thresholds, respectively [29]. These findings indicate that, although the CVS appears to be more frequent in POMS than in MS-mimicking conditions, it may also be observed in other pediatric inflammatory demyelinating disorders. Taken together, while robust evidence supports the diagnostic value of CVS in AOMS, the limited number of POMS studies warrants caution in applying CVS-based diagnostic criteria to POMS and underscores the need for pediatric-specific validation.

Methodological limitations and gaps in the literature

Several limitations of existing literature must be acknowledged. The number of available studies is small, and most cohorts are modest in size, limiting statistical power and precision. Furthermore, multiple included studies were at high risk of bias, particularly with respect to patient selection. Potential participant overlap across studies from the same centers cannot be excluded. MRI acquisition protocols and analytical methods varied widely across studies, complicating cross-study comparisons and contributing to substantial heterogeneity, particularly for CVS. This variability also included inconsistent use of post-contrast susceptibility imaging across studies, which may influence the visibility of susceptibility-based biomarkers, particularly the CVS. Most studies were cross-sectional and retrospective, precluding assessment of lesion evolution and limiting investigation of how lesion dynamics influence the prevalence of PRLs and CVS. This is particularly relevant when assessing conditions such as MOGAD, which exhibits a high rate of lesion resolution over time. In addition, not all studies applied currently recommended NAIMS criteria for PRL assessment, as some predated their formalization. Reporting of inter- and intra-rater agreement was inconsistent, with only a subset of studies providing reliability metrics. Moreover, none of the available studies explicitly evaluated the incremental diagnostic value of PRLs or CVS beyond conventional MRI criteria.

These limitations constrained the robustness of the present meta-analysis. The small number of studies precluded meta-regression or subgroup analyses to explore sources of variability such as MRI field strength, acquisition protocols, disease duration, or analytical approaches. Consequently, pooled estimates—especially for CVS-related outcomes—should be interpreted with caution.

Further variability likely arose from differences in cohort selection, with some studies focusing on patients close to disease onset and others including more heterogeneous or longer disease durations. Finally, many studies lacked comparative pediatric MS-mimicking cohorts, limiting assessment of diagnostic specificity and comparative performance, underscoring the need for future studies incorporating well-characterized control groups within standardized diagnostic frameworks.

Clinical implications and future directions

Taken together, our findings suggest that PRLs are a frequent and relatively consistent imaging feature in POMS. Limited comparative data suggest promising diagnostic specificity, although the available pediatric evidence remains scarce. CVS is also commonly observed in POMS, but exhibits marked variability, highlighting the need for standardized acquisition and analysis protocols and pediatric-specific validation of diagnostic thresholds.

CVS also appears to have more limited specificity, particularly when considering MOGAD. Future research should prioritize prospective multicenter studies using standardized susceptibility-based MRI protocols and standardized scoring criteria to better define the prevalence and diagnostic performance of PRLs and CVS in POMS. Studies including well-characterized pediatric MS-mimicking cohorts such as MOGAD are needed to clarify diagnostic specificity and refine pediatric-specific thresholds. In addition, longitudinal investigations will be important to characterize lesion evolution and determine how disease stage influences the detectability and clinical relevance of susceptibility-based imaging biomarkers in POMS. Overall, the findings of this meta-analysis should be interpreted with caution, as they are based on a limited number of relatively small studies. Nevertheless, they provide the first comprehensive synthesis of the available evidence on susceptibility-based MRI biomarkers in POMS.

At the same time, they emphasize the urgent need for larger, prospective, multicenter studies including relevant MS-mimicking conditions to fully establish their diagnostic value in children and to maximize their potential in clinical decision-making.