Small-sample performance and underlying assumptions of a bootstrap-based inference method for a general analysis of covariance model with possibly heteroskedastic and nonnormal errors

Impact of the Ablative Margin on Local Tumor Progression after Radiofrequency Ablation for Lung Metastases from Colorectal Carcinoma: Supplementary Analysis of a Phase II Trial (MLCSG-0802)

PURPOSE

To explore what extent of ablative margin depicted by computed tomography (CT) immediately after radiofrequency (RF) ablation is required to reduce local tumor progression (LTP) for colorectal cancer (CRC) lung metastases.

MATERIALS AND METHODS

This retrospective study was undertaken as a supplementary analysis of a previous prospective trial. Seventy patients (49 men and 21 women; mean age ± standard deviation, 64.9 years ± 10.6 years) underwent RF ablation for CRC lung metastases, and 95 tumors that were treated in the trial and followed up with CT at least 12 months after RF ablation were evaluated. The mean tumor size was 1.0 cm ± 0.5 cm. The ablative margin was estimated as the shortest distance between the outer edge of the tumor and the surrounding ground-glass opacity on CT obtained immediately after RF ablation. The impact of the ablative margin on LTP was evaluated using logistic regression analysis. Multivariate logistic regression analysis was also performed to identify the risk factors for LTP. The result was validated with multivariate logistic regression applying a bootstrap method (1,000 times resampling).

RESULTS

The mean ablative margin was 2.7 mm ± 1.3 (range, 0.4-7.3 mm). LTP developed in 6 tumors (6%, 6/95) 6-19 months after RF ablation. The LTP rate was significantly higher when the margin was less than 2 mm (P = .023). A margin of <2 mm was also found to be a significant factor for LTP (P = .048) on multivariate analysis and validated using the bootstrap method (P = .025).

CONCLUSIONS

An ablative margin of at least 2 mm is important to reduce LTP after RF ablation for CRC lung metastases.

Analysis of covariance under variance heteroscedasticity in general factorial designs

Adjusting for baseline values and covariates is a recurrent statistical problem in medical science. In particular, variance heteroscedasticity is non-negligible in experimental designs and ignoring it might result in false conclusions. Approximate inference methods are developed to test null hypotheses formulated in terms of adjusted treatment effects and regression parameters in general analysis of covariance designs with arbitrary numbers of factors. Variance homoscedasticity is not assumed. The distributions of the test statistics are approximated using Box-type approximation methods. Extensive simulation studies show that the procedures are particularly suitable when sample sizes are rather small. A real data set illustrates the application of the methods.

A simulation study to compare robust tests for linear mixed-effects meta-regression

The explanation of heterogeneity when synthesizing different studies is an important issue in meta-analysis. Besides including a heterogeneity parameter in the statistical model, it is also important to understand possible causes of between-study heterogeneity. One possibility is to incorporate study-specific covariates in the model that account for between-study variability. This leads to linear mixed-effects meta-regression models. A number of alternative methods have been proposed to estimate the (co)variance of the estimated regression coefficients in these models, which subsequently drives differences in the results of statistical methods. To quantify this, we compare the performance of hypothesis tests for moderator effects based upon different heteroscedasticity consistent covariance matrix estimators and the (untruncated) Knapp-Hartung method in an extensive simulation study. In particular, we investigate type 1 error and power under varying conditions regarding the underlying distributions, heterogeneity, effect sizes, number of independent studies, and their sample sizes. Based upon these results, we give recommendations for suitable inference choices in different scenarios and highlight the danger of using tests regarding the study-specific moderators based on inappropriate covariance estimators.