Investigating the prognostic value of TP53 and PIK3CA mutations in metastatic colorectal cancer (mCRC): Applying Structured Query Language (SQL) algorithms to real-world data from a high-volume U.K. tertiary cancer center.

Abstract

57 Background: Real world data is a valuable research resource but manual data collection is time consuming and labour intensive. As there is no consensus on the prognostic role of TP53 and PIK3CA mutations in mCRC, we assessed the feasibility of using SQL, a programming language, to derive data from routine electronic clinical notes to investigate the prognostic impact of TP53 and PIK3CA mutations in mCRC. Methods: A cohort of patients (pts) diagnosed with metastatic or recurrent colorectal cancer as per ICD-10 classification from January 2015 to December 2017 and managed at the Royal Marsden Hospital were identified using SQL algorithms developed in-house. Baseline demographics, histopathological and molecular characteristics and death dates were derived from the Electronic Patient Record and extracted in a structured format for statistical analysis. Mutational analysis of TP53 and PIK3CA were performed with standard of care KRAS, NRAS and BRAF testing using next generation sequencing. Overall survival (OS) according to TP53 or PIK3CA mutational status was estimated using the Kaplan-Meier method. Uni- and multivariate Cox regression included KRAS, NRAS, BRAF, sidedness and mismatch repair (MMR) status. Association between TP53 or PIK3CA and MMR status was tested by the Chi squared test. Results: A total of 367 mCRC pts were identified; 10 were excluded due to ineligibility or inadequate data availability. Based on a final dataset of 357 pts, 342 and 354 pts had TP53 and PIK3CA results available. The incidence of TP53 mutations was 75% (n = 257/342) and PIK3CA 16% (n = 55/354). Co-mutations with KRAS, NRAS and BRAF were seen (Table). The overall median follow-up was 42.3 and 42.7 months for TP53 and PIK3CA respectively. There was no difference in OS between TP53 mutant (MT) and wild type (WT) pts (22.0 vs. 22.8 months, p = 0.96) and between PIK3CA MT and WT pts (21.7 vs. 22.4 months, p = 0.49). Right sided and BRAF MT tumours were associated with poorer survival than left sided and BRAF WT tumours when all other factors were constant for TP53 (p<0.001 & p=0.033 respectively) and PIK3CA (both p<0.001) OS on multivariate analysis. MMR deficient tumours were significantly more frequent in TP53 WT compared to TP53 MT tumours (15% vs. 4%, p=0.001) while they were significantly more frequent in PIK3CA MT compared to PIK3CA WT tumours (20% vs. 5%, p < 0.001). Conclusions: Bespoke SQL algorithms enables large-scale data extraction to facilitate research. Based on this dataset, we have shown that TP53 and PIK3CA mutations have no prognostic impact in mCRC. [Table: see text]