We appreciate Ferrari et al’s meta-analysis of the no specific molecular profile (NSMP) prognostic markers. However, important methodologic and biologic limitations remained unaddressed, and we highlight key issues requiring further clarification to strengthen the interpretation and future application of their findings.
The article implicitly assumes that NSMP is a homogeneous group suitable for pooled analysis, which may not be valid. NSMP is the most heterogeneous The Cancer Genome Atlas Program subtype. Although the study combines NSMP cohorts from diverse centers, stages, histologies, and treatments, the baseline tables show very high rates of missing information for key pathologic factors, namely myometrial invasion at 71.5% to 79.3%, International Federation of Gynecology and Obstetrics stage at 37.7% to 54.7%, and lymphovascular space invasion (LVSI) at 21.7% to 36.5%. These core prognostic variables are essential for determining whether the L1 cell adhesion molecule (L1CAM-positive) group contained more grade 3 (G3) tumors (≥50% invasion) or LVSI-positive disease or whether the estrogen receptor (ER)-negative subgroup inherently included higher-risk patients. The pathologic distributions also differed across biomarker groups; G3 tumors accounted for 10.9% in the progesterone receptor (PR) group and for 7.5% in the L1CAM group, and stage III to IV disease accounted for 6.6% in the ER group and 9.0% in the L1CAM group. Such discrepancies suggest that the pooled hazard ratios may reflect a baseline risk imbalance rather than true molecular effects, because uncontrolled differences can generate pseudo-significance. Future meta-analyses should evaluate pathologic heterogeneity and apply multivariate pooling, propensity matching, or individual patient data modeling.
In the analysis, L1CAM, ER, and PR were treated as independent prognostic indicators, thereby oversimplifying their biologic interactions. ER positivity reflects an epithelial, low-epithelial-mesenchymal transition (EMT) phenotype, whereas L1CAM is linked to EMT activation, suggesting potential antagonism. Combinations such as ER+/L1CAM+ or ER–/L1CAM+ may therefore show nonadditive risk effects, however, no interaction or subgroup analyses were performed. The baseline subsets also differed with missing age data for 80.3% in the ER group and 72.5% in the L1CAM group, thereby challenging the comparability of independent hazard ratios. In addition, substantial immunohistochemistry (IHC) heterogeneity—antibody clones, scoring systems, and variable cutoffs (1%, 10%, H-score)—was not methodologically stratified. Future studies should incorporate synergistic models, standardized IHC methods, and methodologic stratification. Finally, the article assumes static biomarker status, however, ER may decline and L1CAM increase in recurrent disease. Findings based on primary tumors cannot guide recurrence prediction. Future NSMP studies should use longitudinal sampling and dynamic monitoring, because single-marker hazard ratios are insufficient for modern multidimensional risk assessment.
Ferrari et al offer a valuable foundation, but meaningful NSMP risk stratification requires addressing heterogeneity, IHC variability, biomarker dynamics, and interactions through future prospective, multicenter, high-dimensional studies.
Y.L. and X.C. are co-first authors.
The authors report no conflict of interest. No financial relationships, funding sources, or relevant affiliations influenced the content or preparation of this manuscript.
This study was funded by the Fujian Provincial Science and Technology Innovation Joint Fund Project under grant number 2023Y9454, the Fujian Provincial Natural Science Foundation Project under grant number 2024J011087, and the Fujian Provincial Health Commission Science and Technology Plan Project under grant number.
References
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
