Se 12

Stata SE 12.0 was used for the statistical analysis. Differences in baseline characteristics between the two cohorts were assessed by means of Pearson’s Chi-square tests and independent sample t-tests. Outcomes were proportions of given treatments, cumulative incidences of recurrence and overall mortality by care setting. Cumulative incidences of recurrence were calculated using the Cumulative Incidence Competing Risk (CICR) method considering death without recurrence as a competing event [22 (link), 23 (link)]. Overall mortality was calculated using the Kaplan Meier method. Cox proportional hazard models were used to characterize the influence of care setting on recurrence risk and overall mortality. Covariates were included in the multivariate model if they were judged to be clinically relevant, and comprised age at diagnosis, comorbidity, histological grade, T stage and N stage. Treatments were not included in the multivariable model because one of the mechanisms through which oncogeriatric care influences outcomes is through treatment. All statistical tests were two-sided and a p value of < 0.05 was considered statistically significant.

The pooled HR and its corresponding 95% CI for prognosis were calculated with Stata SE12.0. The combined ORs and its corresponding 95% CI for clinical parameters were calculated by RevMan5.3 software.
Statistical heterogeneity among studies was assessed with I²test and Q statistic test, the fixed-effect model was applied when no obvious heterogeneity was observed among studies, otherwise, the random effects model was applied to calculate parameters when there was significant heterogeneity across studies(I² > 50% or P_h<0.05 suggested significant heterogeneity).
Funnel plots and Begg's test/Egger's test were involved to search the potential publication bias. The sensitivity analysis was also performed to test the reliability of the combined results. A p value less than 0.05 was considered statistically significant.

The present meta-analysis was conducted using the RevMan5.2 software and Stata SE12.0 software. The heterogeneity among the enrolled studies was determined by the chi-square-based Q-test and I² statistics; a P-value for the Q-test <0.05 and a I²-value>50% were considered as indicators of severe heterogeneity. The random-effects model was applied to the studies with a significant heterogeneity (P_Q≤0.05, I²≥50%). Otherwise, the fixed-effects model was adopted (P_Q>0.05, I²<50%). Potential publication bias was assessed using a funnel plot, and the sensitivity analysis was also performed to ensure the reliability of the results. The P-value <0.05 was considered statistically significant.

We used Stata SE 12.0 analysis software for the statistical analyses of the selected studies. The relative risk (odds ratio (OR)) was used to analyse the statistical data, and the weighted mean difference was used to analyse the results. The effects were expressed using 95% confidence intervals (CIs). First, we tested for heterogeneity of the included studies using the chi-square test. When studies in the group were not heterogeneous (P ≥ 0.1, I² ≤ 50%), we used the fixed-effects model for the meta-analysis. When heterogeneity (P < 0.1, I² > 50%) was present, we analysed the reason and used sensitivity analyses to process the data before excluding studies of lower quality and evaluating the stability of the results of the meta-analysis. The studies whose heterogeneity still could not be eliminated were consolidated with a random-effects model. We used a forest map to list the results and an intention-to-treat analysis to determine whether there was attrition bias. However, if there was clear clinical heterogeneity among the studies, we did not merge them but rather performed a descriptive qualitative analysis. The funnel plot test was used to examine the existence of publication bias.

Stata SE 12.0 was used to conduct our systematical review, making forest plots and describe publication bias. HR and confidence interval (CI) were used as effect sizes. If the P value was less than 0.05, the difference between two arms had a statistical significance. The heterogeneity would be low, moderated and high, if the I² value was less than 25%, 25–50% and over 50%, respectively. In this analysis, the null hypothesis that the studies were homogenous would be rejected if P for heterogeneity was less than 0.10 or I² > 50%. When there was significant heterogeneity among the results of included study, the random-effects model was used to calculate summary estimate [14 (link)]. Otherwise, the summary estimate was calculated based on the fixed effects model, reported using the inverse variance method, assuming that the studies included in the meta-analysis had the same effect size.