Document resource
Background Lymphocyte activation gene-3 (LAG-3) is an inhibitory checkpoint receptor upregulated on chronically stimulated (‘exhausted’) T-cells within the tumor microenvironment. Although immune checkpoint inhibitors (ICIs) targeting the PD-1/PD-L1 axis prolong survival in metastatic urothelial carcinoma, their efficacy may be limited by persistent LAG-3-mediated T-cell exhaustion. We applied an AI-driven quantitative multiplex immunofluorescence (QMiF) pipeline to calculate the ratio of LAG-3 + immune cells to tumor cells and assessed its association with progression-free survival (PFS) and overall survival (OS) in stage IV metastatic urothelial cancer patients treated with multiple lines of immunotherapy (durvalumab, pembrolizumab, atezolizumab).Methods In our single-center retrospective cohort (n=57), primary tumor biopsies stained for DAPI, LAG-3, CK, and TROP2 were imaged by QMiF, and an AI pipeline quantified LAG-3 + and tumor cells (figure 1). We compared continuous LAG-3: tumor ratios between responders (n=30) vs non-responders (n=27) and non-hyperprogressors (non-HP, n=35) vs hyperprogressors (HP, n=22) defined by RECIST criteria, using Mann-Whitney U tests and Cohen’s d, then dichotomized at the 65th percentile. PFS and OS were assessed by Kaplan-Meier and univariate Cox analyses, with multivariable Cox models adjusted for age, sex, BMI, ECOG status, and ICI cycle count.Results Responders exhibited a lower LAG-3 +: tumor cell ratio than non-responders (p < 0.001; Cohen’s d = 1.00; figure 2A), and non-HP had a lower ratio than HP (p = 0.038; d = 0.65; figure 2B). The ratio distributions in the low- and high-risk groups were markedly distinct (p < 0.0001; d = 2.12; figure 2D). At one year, 85% of high-ratio versus 65% of low-ratio patients had progressed. In univariate Cox analysis, a high ratio was associated with shorter PFS (HR 2.07 [1.14–3.72]; p = 0.016; c-index 0.59; figure 2C) and shorter OS (HR 1.49 [0.84–2.65]; p = 0.26). In multivariable analysis adjusting for other clinical parameters, a high LAG-3: tumor ratio (HR 1.90 [1.02–3.73]; p = 0.045) and fewer ICI cycles (HR 0.90 per cycle [0.87–0.95]; p < 0.001) remained independent predictors of shorter PFS while the ECOG status was borderline significant (HR 1.48 [1.0-2.20]; p=0.051).Conclusions AI-driven quantification of the LAG-3: tumor cell ratio by QMiF represents a promising prognostic biomarker for PFS in metastatic urothelial carcinoma treated with PD-1/PD-L1 inhibitors. Given the modest sample size and single-center design, these findings require external validation and evaluation on whether this marker is predictive of benefit of immunotherapy.Ethics Approval Ethical approval and informed consent were obtained in accordance with Emory University and Cleveland clinic guidelines (STUDY0000535).Abstract 92 Figure 1QMiF whole slide images and respective ROIs showing the detected LAG-3+ cells (red circles) by our model alongside CK and TROP2 staining in a responder (A) versus a non-responder (B). The cell boundaries are highlighted in greenAbstract 92 Figure 2Panels A-B: violin plots of LAG-3+ to tumor cell ratios by response group (A) and hyperprogessor status (B); Panel C: 5-year PFS Kaplan-Meier curves for low versus high ratio; Panel D: Violin plot of LAG-3+: tumor cell ratio distribution by risk