Background
Patients with advanced lung cancer represent a heterogenous population with varying patterns of metastasis. Those with liver metastases may represent a unique cohort with differential response to therapy, including immunotherapy in NSCLC. Novel Natural Language Processing (NLP) and Artificial Intelligence (AI) technology enables automated extraction of real-world data to examine these populations at greater scale than current manual chart abstraction processes, helping clinicians make more informed treatment decisions.
Method
Patients diagnosed with stage IIIB/IV lung cancer who received first-line systemic therapy at the Princess Margaret Cancer Centre between 2015 and 2018 were reviewed using the DARWEN™ NLP and AI data abstraction platform developed by Pentavere. Data extracted include tumour histology, patient age, sex, ECOG performance status, smoking status, biomarker status, PD-L1 expression, sites of metastases, treatment details and survival.
Result
Of 615 patients with accessible electronic pathology records, 540 (87.8%) had NSCLC and 333 (54.1%) received systemic therapy. In those patients treated with first-line therapy (immunotherapy 10.2%, targeted therapy 30.9%, chemotherapy 62.7%), 27.3% (91/333) had liver metastasis at any point from baseline to end of follow up (median follow up 8 months). 280 patients had NSCLC and received systemic therapy and were included in subsequent analysis. Of these, 69 (24.6%) had liver metastases at any point and overall survival was worse in those patients 544 vs 715 days (p=0.006). Liver metastases were more commonly seen in those with more metastatic sites (OR: 1.42, 95% CI: 1.19-1.70, p = <0.001). By contrast, those with EGFR mutant lung cancer were less likely to develop liver metastasis (OR: 0.45, 95% CI: 0.23-0.87, p=0.02). Using Cox regression analyses, after controlling for age, sex, baseline performance status, baseline smoking status, first line treatment, total number of metastatic sites and baseline LDH, presence of liver metastasis remained significantly associated with worse survival (HR: 1.78, 95% CI: 1.14-2.76, p=0.01). Elevated baseline LDH, a known poor prognostic factor, was also associated with worse overall survival (HR: 1.58, 95% CI: 10.6-2.35), p=0.02). No differential effect by type of therapy was seen.
Conclusion
The presence of liver metastases confers worse prognosis in advanced non-small cell lung cancer patients. This effect was observed irrespective of treatment type and highlights the need for additional treatment options which are efficacious in this patient population. Larger cohort studies may help identify patients with liver metastases that may benefit from specific therapeutic strategies in the future. NLP and AI technologies like DARWEN™ can rapidly generate population-based datasets and provide clinicians with timely access to previously unavailable information on treatment patterns and outcomes which can lead to improved care.
