Prognostic value of microRNA-21 in intra- and extrahepatic cholangiocarcinoma after radical resection: cohort study

Résumé fourni par la source

The clinical outcomes of patients with cholangiocarcinoma are poor, with a median survival of 17–38 months after resection, depending on the anatomical location of the primary tumour and the presence of high-risk features (for example positive resection margins and lymph node metastases)1,2. Incorporating biomarkers such as microRNAs into prognostic models for cholangiocarcinoma might improve patient risk assessment, enable personalized clinical decision-making and provide more precise prognosis estimates. A previous meta-analysis identified microRNA-21 (miR-21) as an ‘ideal prognostic marker for clinical decision-making’ but it did not compare its performance with commonly used clinicopathological variables such as resection margin and tumour grade. Previous studies revealed that the prognostic value of a novel biomarker can be substantially reduced after correcting for commonly measured prognostic variables3–5. The aim of this study was to assess the association between miRNA-21 and overall survival (OS) in cholangiocarcinoma and assess its performance in prognostic prediction models. Data from patients who underwent curative-intent resection for cholangiocarcinoma between July 2002 and July 2021 at the University of Bologna, University of Pisa, and Amsterdam University Medical Center were extracted from electronic medical records. Adjuvant chemotherapy with gemcitabine was administered to all patients from the University of Bologna and Pisa, but not to those from the Amsterdam University Medical Center. The study was approved by the institutional review board of each participating centre. Details regarding the RNA extraction, expression analyses and data normalization are provided in the Supplementary materials. Total RNA was isolated from 10 μm thick formalin-fixed paraffin-embedded tissue (FFPE) tumour sections. RNA was used for expression analysis of miR-21 by quantitative real-time PCR (qPCR). RNA (10–100 ng) was reverse transcribed and the resulting cDNA was amplified using the specific Taqman MicroRNA assays (Life Technologies) for miR-21 and RNU6B (assay ID, 000397 and 001093 respectively). Continuous and categorical baseline variables were reported as median (interquartile range) or as numbers and percentages respectively. Cox regression modelling was used to assess the association between miR-21 and OS, and to assess the added value of miR-21 to routinely available prognostic markers in a multivariable model. In the multivariable model, the following prespecified prognostic covariates were included: miR-21, age, sex, lymphatic invasion, vascular invasion, perineural invasion, resection margin, T stage and N stage. Age and miR-21 were modelled using restricted cubic splines6. Missing data were handled using multiple imputation (60 imputations)7. The imputation model included the event variable, the Nelson–Aalen estimate of the cumulative baseline hazard and all prognostic covariates included in the full Cox regression model8. A P value lower than 0.05 was considered statistically significant. All statistical analyses were performed in R, version 4.2.1 (R Foundation for Statistical Computing), and Stata, version 17.0 (StataCorp). Details regarding the statistical analyses are provided in the Supplementary materials. All patients gave informed consent. The protocol was approved by the University of Bologna Ethics Committee. All methods were carried out in accordance with relevant guidelines and regulations. A total of 131 extrahepatic and intrahepatic cholangiocarcinoma patients aged between 34 and 82 years were included. Baseline characteristics are shown in Table 1 and the frequency of missing data is shown per variable in Table S1. Baseline characteristics Values are n (%) unless otherwise indicated. i.q.r., interquartile range; miR-21, microRNA-21. Baseline characteristics Values are n (%) unless otherwise indicated. i.q.r., interquartile range; miR-21, microRNA-21. OS in three equally sized groups of patients with low, moderate and high miRNA-21 expression is shown in Fig. 1a. Compared with the group of patients with high miR-21 expression, moderate miR-21 expression and low miR-21 expression were associated with significantly longer OS (hazard ratio (HR) 0.47 (95% c.i. 0.28 to 0.79) and 0.49 (0.24 to 0.66) respectively). This association was similar between men and women (interaction test, P = 0.422), and remained consistent across age groups (interaction test, P = 0.255). Prognostic value of miRNA-21 in cholangiocarcinoma. a, Kaplan–Meier curve for three equally sized groups of patients with low, moderate and high expression of miRNA-21; b, adjusted (blue) and unadjusted (black) association between miR-21 and overall survival; c, importance of each prognostic variable in the full Cox regression model, as measured by the partial Wald χ2 minus the predictor degrees of freedom. Higher χ2 values indicate higher prognostic value. a, low miR-21, Ct cycle 24.6 to 30 (9 ICC, 34 ECC); moderate miR-21, Ct cycle 22.9 to 24.6 (9 ICC, 35 ECC); high miR-21, Ct cycle 15 to 22.9 (8 ICC, 34 ECC); b, miR-21, number of Ct cycles for miR-21; c, Wald χ2 values are derived from the fully adjusted Cox regression model, after multiple imputation. ICC, intrahepatic cholangiocarcinoma; ECC, extrahepatic cholangiocarcinoma; Ct, cycle threshold; miR-21, microRNA-21. In univariable analysis, miR-21 was strongly associated with shorter OS (standardized HR per standard deviation increase in miR-21 2.13 (95% c.i. 1.56 to 2.94); P < 0.0001). The association between miR-21 and OS remained consistent (adjusted standardized HR 2.08 (95% c.i. 1.54 to 2.86); P < 0.0001; Fig. 1b), after correcting for conventional clinicopathological variables (that is age, sex, vascular invasion, perineural invasion, lymphatic invasion, resection margin, T stage and N stage). This association was similar between patients with intrahepatic versus extrahepatic cholangiocarcinoma and remained consistent in four sensitivity analyses (Supplementary results). In the fully adjusted Cox regression model, the prognostic value of miR-21 was higher than all conventional clinicopathological variables (Fig. 1c), and miR-21 had significant prognostic value after correcting for these variables (likelihood ratio test, P < 0.0001). In the fully adjusted model, miR-21 contained 70% of the prognostic information provided by all clinical, pathological and biomarker variables combined (likelihood ratio χ2 before and after adding miR-21, 21.4 versus 66.7). Prognostic performance increased substantially when adding miR-21 to a Cox regression model including all conventional clinicopathological variables (Harrell’s C-statistic, 0.76 versus 0.66; difference in C-statistic, 0.09 (95% c.i. 0.04 to 0.14); P = 0.0002). The incremental prognostic value of miR-21 remained similar after correcting for overfitting (difference in overoptimism-corrected C-statistic, 0.10). In this cohort study of 131 patients with intrahepatic or extrahepatic cholangiocarcinoma, miR-21 expression was strongly and independently associated with OS allowing patient stratification into distinct prognostic risk groups and adding prognostic value over conventional clinicopathological variables. The association between miR-21 and OS was consistent in direction and magnitude between men and women, across age groups and across several sensitivity analyses. In a recent meta-analysis, miR-21 has been described as a potential prognostic biomarker for cholangiocarcinoma. The methodology of the current analysis strengthens these results9–13. In contrast to previous analyses, miR-21 was not dichotomized at an arbitrary cut off (for example the median)7,14,15; the multiple imputation approach used avoided discarding patients with missing data, and allowed a higher statistical accuracy; the study was not focused on the crude unadjusted association between high versus low miR-21 expression and OS but miR-21 performance was adjusted for known prognostic co

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.