Page 133 - EJMO-9-3

Eurasian Journal of
Medicine and Oncology KRAS TP53 cholangiocarcinoma

tests were employed: the Q-test and the I statistic. The LLC, USA), and R version 4.2.3 (The R Foundation for
2
Q-test tests the null hypothesis that all included studies Statistical Computing, Austria). These tools facilitated
share the same effect size. A p>0.05 indicates no statistically rigorous and transparent analysis of the data, ensuring the
significant heterogeneity, suggesting consistency among reliability and accuracy.
study findings, whereas a p=0.05 or less indicates significant
2
heterogeneity. I statistic quantifies the percentage of total 3. Results
variation across studies that is due to heterogeneity rather 3.1. Study selection and characteristics
than chance. An I value >50% is typically interpreted
2
as moderate-to-high heterogeneity. In the presence of This meta-analysis identified a total of 458 articles through
significant heterogeneity, defined by either a Q-test p≤0.05 systematic database searches, aimed at examining the
or an I value exceeding 50%, a random-effects model was relationship between KRAS and TP53 gene mutations and
2
applied for the meta-analysis. This model assumes that the survival prognosis in cholangiocarcinoma patients. After
true effect size varies between studies, accounting for both an initial screening of titles and abstracts, 76 articles were
within-study and between-study variation. In contrast, excluded due to duplication, and 324 articles were excluded
when I was <50%, indicating low heterogeneity, a fixed- because they did not meet the inclusion criteria based
2
effects model was used, which assumes a common effect on the information available in their titles and abstracts.
size across all studies and provides a more precise estimate Subsequently, a more detailed assessment of the full-text
under such conditions. articles led to the exclusion of an additional 18 articles due
to insufficient data, which prevented them from contributing
To further evaluate the robustness of the results, a
sensitivity analysis was performed. In this process, each meaningfully to the analysis. Furthermore, 27 articles were
excluded because their subject matter did not align with
study was sequentially removed from the analysis to the focus of this meta-analysis, which specifically addressed
observe the effect on the overall results. This helps identify the prognostic effects of KRAS and TP53 mutations in
any studies that may disproportionately influence the cholangiocarcinoma. In addition, one article involving a
findings and allows for a more comprehensive assessment clinical trial was excluded as clinical trials typically do not
of the reliability of the meta-analysis conclusions. In provide the same type of observational data necessary for
addition, different statistical models were applied to meta-analysis. Ultimately, 12 studies were deemed eligible
assess whether the results were consistent across various for inclusion, involving a total of 1,126 patients. Among
methodological approaches, thus strengthening the these, 167 patients had KRAS mutations and 176 patients had
validity of the conclusions. TP53 mutations, whereas 634 patients had TP53 wild-type

Publication bias is a critical concern in meta-analyses, and 816 patients had KRAS wild-type. The studies included
as studies with significant or positive results are more in the analysis were all retrospective cohort studies, reflecting
likely to be published, leading to an overestimation of the a diverse array of patient populations and clinical settings.
true effect size. To evaluate publication bias, both Egger’s Retrospective cohort studies are valuable for understanding
and Begg’s tests were employed. These tests assess the long-term outcomes and identifying associations between
symmetry of the funnel plot – a graphical representation genetic mutations and survival in real-world clinical practice.
of the relationship between study precision (typically the Regarding the geographic representation of the studies,
standard error) and effect size. A non-significant p-value six studies (50%) focused on Asian populations, whereas
(≥0.05) in either test suggests that there is no substantial the remaining focused on European and American
publication bias, whereas a p<0.05 indicates potential bias, populations. Among the included studies, three involved
possibly due to unpublished studies with negative or non- patients who had received immunotherapy as part of their
significant results, small sample sizes, or extreme findings. treatment regimen. Five of the studies focused specifically
To correct for potential publication bias, the trim-and-fill on patients with intrahepatic cholangiocarcinoma (ICC). In
method was applied. This statistical technique estimates contrast, other studies included patients with extrahepatic
the number of potentially missing studies that may be (ECC) or hilar cholangiocarcinoma. Three studies involved
causing asymmetry in the funnel plot. It then adjusts the patients who were receiving non-surgical treatments,
overall effect size by imputing these missing studies and and two studies included patients with advanced-stage
recalculating the pooled estimate. This approach helps to disease (stage III and IV), which is associated with poorer
correct for bias introduced by selective publication and outcomes. The inclusion of patients across different
enhances the robustness of the meta-analysis conclusions. treatment modalities and disease stages contributed to the
All statistical analyses were carried out using RevMan comprehensiveness of the analysis, allowing for a broader
5.4 (The Cochrane Collaboration, UK), Stata 17 (StataCorp representation of clinical scenarios.

Volume 9 Issue 3 (2025) 125 doi: 10.36922/EJMO025120063