An analysis of findings from multiple studies confirmed the prognostic efficacy of tumor-infiltrating lymphocytes (TILs) in patients with melanoma and found that electronic total TILs (etTILs) outperformed automated TIL% score, calculated by the machine learning algorithm NN192. etTILs also have potential in predicting risk of recurrence in subgroups of patients which could help select patients for adjuvant immunotherapy.

Automated TIL% score, Thazin Nwe Aung, PhD, Yale School of Medicine, New Haven, Connecticut, and co-authors wrote, “has been shown to be prognostic in melanoma but its clinical utility has not yet been broadly proven … The use of NN192 machine learning algorithm could be a valuable and easy-to-implement tool for prospective testing of patients with early-stage melanoma and could be validated as a selector for patients that can safely omit immunotherapy in the adjuvant setting.”

The primary goal of the study was to determine the best practice for utilizing TIL infiltrates to predict disease outcome in melanoma patients, and to prove the clinical utility of TILs in defining the high-risk subset of melanoma patients that are likely to not require treatment with adjuvant immunotherapy.

Using patient data from 5 independent cohorts (n = 785) and serial tissue sections of the Yale TMA-76 melanoma cohort, with a specific focus on 5 TIL variables known to be associated with overall survival (OS), immunofluorescence and Hematoxylin-and-Eosin (H&E) staining were employed to reveal molecular characteristics of TIL phenotypes and their relationship to survival outcomes. The clinical impact of eTILs was compared against automated TIL%. Their findings indicate etTILs outperformed automated TILs (area under curve [AUC]: 0.793; specificity: 0.627; sensitivity: 0.938 vs AUC: 0.77; specificity: 0.51; sensitivity: 0.938, respectively). The study also found a specific molecular subtype of cells (predominantly CD3+ and CD8+ or CD4+ T cells) that represent TILs.

“In the future, the use of eTILs might be complemented with molecular subtyping of cells for more discriminating analyses. The use of a combined marker signature may be proven to be the best approach to define a subset of patients that will not benefit from immunotherapy or might develop significant toxicities,” Dr Aung and colleagues concluded.

Source:

Aung TN, Shafi S, Wilmott JS, et al. Objective assessment of tumor infiltrating lymphocytes as a prognostic marker in melanoma using machine learning algorithms. EBioMedicine. 2022;82:104143. doi:10.1016/j.ebiom.2022.104143