Modeling of Growth Patterns of Postoperative Meningioma Remnants

Introduction: Complete surgical resection of meningiomas is sometimes unattainable, leaving residual disease at risk of recurrence. The monitoring of remnant progression can be challenging and often relies on long-term imaging surveillance. Further research on the growth kinetics of residual meningioma is needed in order to optimize postoperative management. The aim of this study is to describe the volumetric growth patterns of residual meningioma and to identify factors associated with rapid growth.

Methods: In this retrospective cohort study, we included adult patients with a postoperative finding of residual meningioma on imaging between 2010 and 2021. Only WHO grade 1 meningiomas were included. Volumes were measured manually by two independent observers using three-dimensional segmentation software on contrast enhanced T1-weighted MRI. Tumor growth rates (TGR) were derived from volumetric data over time and potential factors associated with rapid growth were investigated. Goodness of fit was compared across linear, exponential, Gompertzian and power models. The ability of the four models to predict future untrained datapoints was assessed.

Results: A total of 61 remnants in 59 patients were followed for an average of 71.1 months postoperatively. Volumetric data was obtained from 500 MRI scans. Mean preoperative tumor volume was 30.01 cm³ and mean initial remnant volume was 2.51 cm³. Mean TGR from furthest timepoints was 1.79% month preoperatively and 1.68% month postoperatively. TGRs normalized toward smaller values over time ([Fig. 1]).

Fig. 1 Remnant TGRs (% month) over postoperative time. Negative TGRs represent spontaneous decrease in remnant volume.

Clustering analysis reduced remnants into 18 rapid growing and 38 non–rapid-growing lesions ([Fig. 2]). Mean TGR within the rapid group was 2.71% month versus 0.20% month in the non–rapid group (p < 0.001).

Fig. 2 Standardized remnant volume over postoperative time (rapid vs. non-rapid clusters).

Univariate analysis revealed young age (p = 0.005), edema (p = 0.044), and small postoperative volumes (p = 0.039) as predictors of rapid growth. Edema was the only factor registering significance on multivariate analysis (OR = 5.816, 95% CI [1.119–30.229], p = 0.036). There was a significant positive correlation between preoperative and postoperative TGRs (Pearson’s r = 0.767, p < 0.001). Linear and exponential models best described residual disease growth behavior ([Fig. 3]).

Fig. 3 Model fitting of two remnants and respective best fits per Akaike information criterion.

Mean percentage prediction error of all four models on untrained datapoints was 17.64% (SD = 18.05) between 0 and 3 months after the last fitted datapoint. Prediction error grew to a mean of 66.12% (SD = 75.89) after 18 months.

Conclusion: This study highlights the significant discrepancy amongst the growth behavior of residual meningioma and identified factors associated with rapid growth. While meningiomas have been described to follow Gompertzian growth curves preoperatively, the progression of remnants in the current study was better described using linear or exponential models. Nevertheless, long-term predictive ability of models on untrained datapoints remains poor. These findings can help guide postoperative management and help in identifying patients at high risk of progression.

 

MissingLinkText

Die Autoren geben an, dass kein Interessenkonflikt besteht.

Publikationsverlauf

Artikel online veröffentlicht:
07. Februar 2025