Data analysis
Primary outcomes (OS, PFS) were analyzed by pooling log hazard ratio
(HR) and standard error using generic inverse variance. These values are
calculated from HR and 95% confidence interval (CI) of multivariable
Cox regression models using Cochrane approved formulas (12). Secondary
dichotomous outcomes (intraoperative complications, postoperative
complications) were expressed as pooled odds ratios and 95% CI. Effect
size of continuous data are expressed as weighted mean difference (WMD)
and 95% CI using means and standard deviation. Data presented as
medians and ranges were converted to mean and standard deviation using
Hozo’s formula (13). Random effect model was opted due to anticipated
heterogeneity.
MOGGE Meta-analysis Matrix (MMM) is a novel illustration tool that is
used to present all possible subgroup analyses. Because of
heterogenicity of study cohorts, a matrix was created using 3 axes:
disease stage, type of chemotherapy (adjuvant and neoadjuvant), and
control group (selLND, noLND). “Disease stage” axis is divided into
columns based on grouped studies covering the same range of stages.
These columns are further transected by “type of chemotherapy” and
“control group”. Thus, each cell of this matrix presents a subgroup
analysis of a group of studies that correspond to the stage above,
chemotherapy to the left and control group to the right. Each cell
contains the pooled HR and 95% CI, number of studies and patients in
each subgroup analysis. The right column presents the total for each
row.