A meta-analysis of models for predicting the incidence of brain metastasis in patients with lung cancer

Background: Brain metastasis (BM) represents a significant clinical challenge in the management of lung cancer, impacting patient prognosis and treatment decisions. While various predictive models have been developed to assess the risk of BM development, their performance and applicability remain under scrutiny. Understanding the accuracy and reliability of these models is crucial for informing clinical practice and improving patient care. We sought to evaluate the predictive ability of existing models predicting BM incidence in lung cancer patients, providing valuable insights into their clinical utility and areas for further refinement.

Methods: Two reviewers conducted a systematic search of numerous databases, including PubMed, EMBASE, Cochrane Library, and CINAHL, and then extracted the relevant information from the eligible studies. For this study, we conducted a meta-analysis using a random-effects model. Pooled estimates were calculated for any combination of outcomes and population when at least two studies had relevant data. Study heterogeneity was assessed using I² statistics. The area under the curve (AUC) of the best model and 95% confidence intervals were set as effect measures. Additionally, funnel plots and results from meta-regression analysis were reported. Meta-analysis was performed with R software ver. 4.3.1, and quality evaluation of evidence was conducted according to the GRADE standard.Results:A total of seven studies with 535,418 participants were eligible. A random-effects model showed that models that could predict the risk of developing BM among lung cancer cases had a pooled AUC of 81% (95% CI: 74% to 87%). However, the studies had significantly higher heterogeneity, with an I² = 99% (p<0.01). Results from meta-regression analysis, accounting for the year of publication (?=-0.013), sample size (?< 0.00), number of variables (?=-0.015), and latitude (?=0.0024) did not show any statistically significant associations with a p>0.05. Egger’s regression test for publication bias did not indicate small study effects (Bias estimate: 4.91, SE = 6.01, p< 0.44).

Conclusions: Our meta-analysis underscores the utility of existing models in predicting the incidence of BM in LC patients, with a pooled AUC of 81%. Further research aimed at refining predictive models and addressing sources of heterogeneity is imperative to enhance clinical decision-making and improve patient outcomes in this population.

Study	AUC [95% CI]	%W
Zhao et al. (2023)	0.7190 [0.5950; 0.9880]	6.9
Li et al. (2022)	0.7670 [0.7120; 0.8230]	15.3
Zuo et al. [a] (2020)	0.7310 [0.7260; 0.7360]	17.1
Zuo et al. [b] (2020)	0.8800 [0.8700; 0.8900]	17.1
Fu et al. (2020)	0.7240 [0.6800; 0.7500]	16.4
Chong et al. (2021)	0.9210 [0.8940; 0.9480]	16.7
Cai et al. (2020)	0.8630 [0.7340; 0.9920]	10.5
Random effects model	0.8630 [0.7388; 0.8722]	z = 23.67; P< 0.0001; I2= 99.3%

 

This material on this page is ©2024 American Society of Clinical Oncology, all rights reserved. Licensing available upon request. For more information, please contact [email protected]