Title: Expression status and prognostic value of PD-L1, FOXP3, CD8 and Ki67 in upper tract urothelial carcinoma

Authors: Jen Liu1; Diana Taheri1; Alcides Chaux2; George J. Netto3

Affiliations: 1Department of Pathology, The Johns Hopkins Medical Institutions, Baltimore, MD; 2Centro para el Desarrollo de la Investigación Científica (CEDIC), Asunción, Paraguay; 3Department of Pathology, The University of Alabama at Birmingham, Birmingham, AL

Last update: 2017-01-27

Methodology

The analysis of PDL1, FOXP3 and CD8/ki67 expression was carried out at the TMA level. The dataset includes 99 patients with upper tract urothelial carcinoma, from whom 297 TMA spots were build from their corresponding tumors.

Evaluation of expression

For each marker, the methodology for evaluation of the immunohistochemical expression was as follows:

  • PDL1: PDL1 was evaluated in tumor and stromal cells, noting the percentage of positive cells.
  • FOXP3, CD8, and Ki67: The expression of these markers was evaluated in lymphocytes infiltrating tumor and stroma, counting the number of positive lymphocytes.

In all cases, the median value of all TMA spots was selected to summarize the expression levels of the marker under evaluation. Thus, for each patient, we had a median value per marker across all sampled tissues.

 

Data analysis

Data analysis was carried out using 3 approaches: descriptive analysis, association analysis, and outcome analysis. Data was analyzed and plots were generated using R version 3.3.2 (2016-10-31) from the R Foundation for Statistical Computing (Vienna, Austria). R packages from the tidyverse were extensively used.

 

Descriptive analysis

Categorical variables were described using frequency tables and barplots. Numerical variables were described using measurements of central tendency and dispersion, histograms, and density plots.

 

Association analysis

Marker values were compared considering clinical, pathologic, and outcome features. In this context, marker values were considered as the outcome variables and the aforementioned features as the predictor variables. Variables were described using measurements of central tendency and dispersion, boxplots, and density plots.

The association was evaluated using either the Mann-Whitney U test for the sum of ranks or the Kruskal-Wallis rank sum test, depending on the features of the predictor variables.

 

Outcome analysis

The outcome analysis included regression modeling and time-to-event (survival) analysis. Markers levels were categorized as high/kow expression using the median as the cutoff point. Outcome variables included bladder recurrence, tumor progression, overall mortality, and cancer-related mortality.

Odds ratios were estimated using unconditional binary logistic regression. Hazard ratios were estimated using Cox’s proportional hazards regression. Survival curves were built using the Kapplan-Meier estimator and compared using the log-rank test.