Results

The search strategy identified 390 records. After title and abstract screening, 14 reports underwent full-text review, of which 11 were included in the quantitative synthesis (Fig. 1) [15–25]. Three studies were excluded for not addressing the review question (n=1), focusing on a non-POMS population (n=1), or reporting a single patient (n=1). In total, 356 patients with POMS were included across the selected studies. The main population characteristics are summarized in Table 1 with supplementary details provided in eTable 3.

Study | Country | Multicentric | Assessing CVS | Assessing PRL | POMS (n) | Age (years) | Female (%) | Disease duration (years) | EDSS | With OCB (%) | Treated with DMTs (%) | Mimics (n) | Matched AOMS (n)
Micheletti 2022 | Argentina | No | No | Yes | 13 | 15 (SD: 2.5) {range: 9;18} | 85% | NA | NA | NA | NA | 16 | –
Boccia 2023 | Italy | No | Yes | No | 10 | 13.56 (SD: 1.79) | 60% | 1.6 (SD: 1.7) | *1.0 [IQR: 1.0;2.0] | 70% | 100% | – | 12
Harrison 2023 | USA | No | Yes | No | 10 | 11.4 (SD: 4.5) | 60% | {range: 0;2.5} | *1.3 | 100% | NA | 10 | –
Margoni 2023 | Italy | No | No | Yes | 13 | *16.2 [IQR: 12.6;16.8] | 85% | *1.0 [IQR: 0.4;1.9] | *1.5 [IQR: 1.0;1.5] | 100% | 100% | – | –
Sacco 2023 | USA | No | Yes | Yes | 26 | *14 [IQR: 13;17] {range: 3;17} | 63% | *0.1 [IQR: 0.0;0.2] {range: 0;3.1} | *1.5 [IQR: 1.0;2.0] {range: 0;2} | 85% | NA | 14 | –
Boccia 2024 | Italy | No | No | Yes | 11 | 16.3 (SD: 2.2) | 64% | 2.4 (SD: 1.5) | *1.0 | 91% | 100% | – | –
de Deus Vieira 2024 | Brazil | No | No | Yes | 10 | *16 | 80% | 2.2 (SD: 1.1) | 1.9 (SD: 1.3) | NA | NA | – | –
Margoni 2024 | Italy | No | Yes | Yes | 22 | *16.4 [IQR: 13.0;17.4] | 86% | *0.7 [IQR: 0.4;1.3] | *1.0 [IQR: 1.0;2.0] | 100% | 73% | – | –
Menascu 2024 | Israel, Czech Republic | Yes | Yes | No | 156 | 14.8 | 60% | NA | **2.0 | NA | NA | – | –
Nistri 2025 | UK | Yes | No | Yes | 54 | 14 (SD: 2.2) | 76% | *0.6 [IQR: 0.2;1.8] | NA | NA | NA | – | –
Kuchling 2025 | Germany | No | Yes | No | 31 | 13.8 (SD: 2.6) | 81% | MRI at first clinical presentation | NA | 97% | NA | – | –

MRI timing relative to clinical presentation or disease activity varied across studies. Some studies performed MRI near the time of first clinical presentation or early in the disease course [19, 25], whereas others included patients during remission phases [18, 22, 24]. In five studies, MRI timing relative to disease activity was not specified [15–17, 21, 23].

MRI acquisition protocols varied substantially across studies, reflecting the use of both research-dedicated and routine clinical scans. Six studies were conducted on 3-T systems [16, 18, 20–23], two on 1.5-T systems [15, 25], two used a combination of 3-T and 1.5-T systems [19, 24], and one study did not report MRI acquisition parameters [17]. Susceptibility-based imaging sequences also varied substantially across studies. Detailed information on MRI protocols and sequences used across studies is provided in eTable 4. Additional details on the PRL and CVS outcomes for each included study are reported in Table 2.

 | Study | NAIMS criteria | Outcome assessed
PRLs | Micheletti 2022 | No* | Presence (≥ 1 PRL)
Margoni 2023 | No** | Presence (≥ 1 PRL)
Sacco 2023 | Yes | Presence (≥ 1 PRL)
Boccia 2024 | Yes | Presence (≥ 1 PRL)
de Deus Vieira 2024 | Yes | Presence (≥ 1 PRL)
Margoni 2024 | Yes | Presence (≥ 1 PRL)
Nistri 2025 | Yes | Presence (≥ 1 PRL)
CVS | Boccia 2023 | Yes | Patient-level proportion of CVS-positive lesions; 40%-CVS rule
Harrison 2023 | Yes | Patient-level proportion of CVS-positive lesions
Sacco 2023 | Yes | Patient-level proportion of CVS-positive lesions
Margoni 2024 | Yes | Patient-level proportion of CVS-positive lesions; 40%-CVS rule; 50%-CVS rule
Menascu 2024 | Yes | 40%-CVS rule
Kuchling 2025 | Yes | Patient-level proportion of CVS-positive lesions; 40%-CVS rule; Select-6 rule

PRLs

Seven studies reporting PRL prevalence in POMS were included [15, 18–22, 24], comprising 149 patients, of whom 109 had ≥1 PRL. Five of seven studies applied the NAIMS criteria for PRL assessment [19, 20–22, 24].

In the random-effects meta-analysis, the pooled proportion of POMS patients with ≥1 PRL was 71.6% (95%CI 61.1–80.2) (Fig. 2). Between-study heterogeneity was low (_I_2=4.8%, _τ_2<0.0001), and the Cochran Q test was not significant (Q=6.30, p=0.39).

Fig. 2: Proportion of patients presenting with PRLs in POMS. Forest plot showing study-specific proportions and the pooled random-effects estimate of the proportion of POMS patients presenting with at least one PRL. Squares represent individual study estimates with size proportional to study weight, horizontal lines indicate 95% confidence intervals, and the diamond represents the pooled estimate. Between-study heterogeneity is reported using I2 and τ2 statistics

Sensitivity analysis using the Freeman–Tukey double arcsine transformation yielded a comparable pooled proportion of 74.1% (95% CI 63.6–83.5), with similarly low heterogeneity (_I_2=13.2%, _τ_2<0.0001; Q=6.91, p=0.33). Leave-one-out analyses showed no disproportionate influence of individual studies, with pooled estimates ranging from 70.1% to 73.2%.

Inter- and intra-rater reliability in PRL assessment

PRL assessment was performed by two raters across all included studies. Good inter-rater reliability was reported in three studies (Cohen’s k=0.75 [15]; ICC=0.75 [22]; inter-rater disagreement in 4.0% of lesions [20]), and good intra-rater reliability was reported in one study (ICC=0.81) [22].

PRL: discrimination between POMS and mimics

Two studies evaluated the performance of PRLs in discriminating POMS from pediatric MS-mimicking conditions. In the study by Sacco et al. [19], none of 14 MOGAD patients showed PRLs, whereas PRLs were detected in 18 of 26 POMS patients, yielding a sensitivity of 69% and a specificity of 100%. In the study by Micheletti et al. [15], none of 16 patients with other pediatric CNS demyelinating disorders showed PRLs, whereas 12 of 13 POMS patients had ≥1 PRL, corresponding to a sensitivity of 92% and a specificity of 100% (95% CI 93.3–100%).

CVS

Six studies reporting CVS data in POMS were included [16, 17, 19, 22, 23, 25]. Five studies (n=99 patients) reported the per-patient proportion of CVS-positive lesions [16, 17, 19, 22, 25], and four (n=219) reported the proportion of patients meeting the 40%-CVS rule [16, 22, 23, 25]. CVS assessment followed the NAIMS criteria in all studies [16, 17, 19, 22, 23, 25].

CVS-positive lesions per patient

In the random-effects meta-analysis, the pooled mean proportion of CVS-positive lesions was 57.9% (95% CI 39.7–76.2) (Fig. 3). Between-study heterogeneity was substantial (_I_2=95.9%, _τ_2=0.021), with significant heterogeneity on the Cochran Q test (Q=97.46, p<0.0001).

Fig. 3: Mean proportion of CVS-positive lesions per patient in POMS. Forest plot showing the study-specific mean proportion of lesions exhibiting the CVS per patient. Squares represent individual study estimates with size proportional to study weight; horizontal lines indicate 95% confidence intervals, and the diamond represents the pooled estimate. Between-study heterogeneity is reported using I2 and τ2 statistics

One study reported rater-specific CVS estimates with poor inter-rater agreement [17]. In the primary analysis, a single study-level estimate was derived by averaging rater-specific means and using the larger standard deviation as a conservative measure of uncertainty. Sensitivity analyses using rater-specific estimates yielded pooled mean CVS proportions of 63.8% (95% CI 42.9-84.8) for rater 1 and 52.0% (95% CI 25.1-78.8) for rater 2, both with very high heterogeneity (_I_2 = 97.1% and 97.7%, respectively; Q-test p < 0.0001). Leave-one-out analyses showed moderate variation in pooled estimates (52.5–63.2%), with persistently high heterogeneity.

40%-CVS rule

In the random-effects meta-analysis, the pooled proportion of patients meeting the 40%-CVS rule was 66.2% (95%CI 10.5–97.0) (Fig. 4), with substantial heterogeneity (_I_2=78.5%, _τ_2=1.887; Q=13.96, p=0.003).

Fig. 4: Proportion of patients meeting the 40%-CVS rule in POMS. Forest plot showing study-specific proportions of patients with POMS meeting the 40%-CVS rule. Squares represent individual study estimates with size proportional to study weight, horizontal lines indicate 95% confidence intervals, and the diamond represents the pooled estimate. Between-study heterogeneity is reported using I2 and τ2 statistics

Sensitivity analysis using the Freeman–Tukey double arcsine transformation yielded a similar pooled proportion of 69.3% (95% CI 13.3-100.0). Leave-one-out analyses indicated instability of the pooled estimate, ranging from 47.1% (95% CI 17.5-78.9) to 78.3% (95% CI 2.3-99.8), consistent with the limited number of studies and pronounced between-study variability.

50%-CVS rule and Select-6 rule

Only one study evaluated the 50%-CVS rule, which was fulfilled by 5 of 22 patients with POMS (23%) [22]. Only one study assessed the Select-6 criterion, reporting fulfillment in all 31 POMS cases analyzed, including four individuals with radiologically isolated syndrome [25].

Inter- and intra-rater reliability in CVS assessment

CVS assessment was performed by two raters in all included studies. Good to excellent inter-rater reliability was reported in three studies (ICC=0.83 [19]; ICC=0.87 [22]; percentage concordance=94.2% [23]), whereas poor inter-rater reliability was reported in one study (ICC=− 0.17 [17]). One study reported high intra-rater reliability (ICC=0.93) [22].

CVS: discrimination between POMS and mimics

Two studies assessed CVS performance in discriminating POMS from disorders on the differential diagnosis of POMS. Harrison et al_._ evaluated 10 POMS and 10 pediatric MOGAD patients and reported higher median percentages of CVS-positive lesions in POMS than in MOGAD (rater 1: 80% vs 9.8%; rater 2: 22.7% vs. 7.5%) [17]. Similarly, Sacco et al_._ found a higher mean proportion of CVS-positive lesions in 26 POMS patients compared with 14 pediatric MOGAD patients (63% vs. 22%) [19].

One study comparing POMS with disease-duration-matched AOMS reported no difference in per-lesion CVS-prevalence (POMS: 52% vs. AOMS: 72%), but a significantly higher proportion of AOMS patients fulfilling the 40%-CVS rule (100% vs. 70%) [16].

Combined PRL and CVS assessment

Only one study evaluated the combined prevalence of PRLs and CVS [22]. The presence of ≥1 PRL or ≥1 CVS-positive lesion was observed in 95% of POMS patients; this proportion decreased to 86% when applying a threshold of ≥1 PRL or ≥3 CVS-positive lesions, and to 77% with a threshold of ≥1 PRL or ≥6 CVS-positive lesions.