Selected Data Analysis Projects
1. Modulation of the anti-tumor immune response and melanoma progression
Our objective was
to investigate the relationship between immune modulation and sentinel lymph
node (SLN) positivity, correlating it with outcome in melanoma patients. I
analyzed the data using Cox model, logistic regression and some parametric and
nonparametric tests. Our analysis
results support an assessment of local immune profiles in both the primary
tumor and the SLN to guide therapeutic decisions. The report on this project was accepted by
the journal Modern Pathology.
2.
Longitudinal lung function analysis in patients enrolled in a treatment program
for World Trade Center related illness
The World Trade Center Environmental Health Center (WTCEHC) is a treatment program for local residents, local workers, and clean-up workers affected by the destruction of the World Trade Center towers on 9/11/2001. We conducted an observational study of longitudinal lung function in patients of the WTCEHC who were clinically followed or monitored. Cross-sectional analysis was performed for baseline visits. Longitudinal evaluation was performed on the population as a whole and subsequently stratified by smoking status, initial spirometry pattern and WTC-related exposure category. Linear mixed-effects modeling was used to investigate the temporal patterns of forced vital capacity (FVC) and forced expiratory volume in one second (FEV1). Improvement in forced expiratory volume and forced expiratory volume in one second was noted for the population as a whole. Heavy smokers did not improve. Spirometry changes differed depending on initial spirometry pattern and exposure category. There are two conference abstracts and one peer-review publication on this project.
3. Volume Dependence of Distal Airway Dysfunction in WTC Exposed Subjects with Normal Airflow
This study was designed to evaluate whether progressive decrease in lung volume would be associated with progressive abnormalities in airway resistance. Subjects were classified into 4 groups to reflect progressive reduction in lung volume. I analyzed the data using linear regression and some nonparametric statistical tests. The analysis result demonstrates that increased impulse oscillometry( IOS) resistance in symptomatic WTC dust exposed subjects was attributable to progressive decrease in resting lung volume. There is one conference abstract on this project and the manuscript of the paper is in preparation.
4. Challenging the current paradigm of melanoma progression: brain metastasis as isolated first visceral site
Melanoma brain metastases that develop as the isolated first visceral site challenge the current paradigm of tumor progression in which brain metastasis is regarded as the final stage. Here we test the hypothesis that melanoma patients who develop brain metastasis as the isolated first visceral site have distinct clinicopathological features at the time of primary melanoma diagnosis. Cutaneous melanoma patients enrolled in two prospectively collected databases were studied (Cohort 1: 1972-1982, Cohort 2: 2002-2009). Patients who developed brain metastasis as isolated first visceral site were compared to: 1) all other patients, 2) patients who developed visceral metastasis: extracranial only or extracranial and brain, and 3) patients who progressed to other isolated visceral sites first. I compared on clinical variables among these subgroups in each cohort using statistical tests, and then combined the p value across the cohorts using Fisher’s method. I analyzed the survival data using Gray’s test and the Fine-Gray model, which account for competing risks.
5. French Health Care Project
The principal aim of this study is to compare access to primary and specialty care in three metropolitan regions of France: Ile de France (IDF), Nord-Pas-de-Calais (NPC) and Provence-Alpes-Côte d’Azur (PACA). To identify the factors that contribute to variations in access to care within and among these regions, I analyzed this large dataset using generalized linear mixed effect model.
6.
Mitotic Index Project
The aim of this study is to assess MPM-2 and PHH3 as more objective means of calculating mitotic rate compared to conventional H&E-based mitotic rate. I evaluated the association between conventional light microscopic evaluation and computer-assisted PHH-3 and MPM-2 analysis using the Pearson correlation coefficient. I evaluated the association between progression-free survival and mitotic using the Cox regression model and log rank test. Our results suggest that PHH3-derived mitotic rate might be useful in predicting worse clinical outcome in melanoma patients with thicker primaries.
7. Melanoma Risk Index Project
The aim of this project is to find out variables that are associated with melanoma from the population-based Minnesota Skin Health Study and to investigate whether these variables could improve the predictive ability of a clinical risk model. I identified the risk factors using association test, Chi-square test and logistic regression model. In addition to the previous reported conventional risk factors (i.e. age; gender; hair, skin, and eye color; mole count, freckling, and family melanoma history), melanocortin receptor (MC1R) polymorphisms, indoor and outdoor ultraviolet light (UV) exposure were also identified to be associated with melanoma risk. We then assessed the predictive ability of the model incorporating both conventional variables and the newly identified risk factors and the model only incorporating the conventional variables by evaluating and comparing AUC (area under the ROC curve). We found that a commonly used method for AUC comparison (Delong’s method) is too liberal. Then we reviewed the literature and found that other AUC comparison methods also had deficiencies. Thus, we proposed a new test for AUC comparison, which is robust. Our analysis demonstrates that the addition of genetics and indoor UV exposure results in a measurable increase in predictive ability when compared to models comprised only of clinical and outdoor UV exposure variables.
8. Meta-analysis and multivariate analysis of melanoma risk factors.
To identify patients at increased risk of developing brain metastasis early, we examined the predictive ability of demographic/primary melanoma characteristics as risk factors for brain metastasis development. Studies with available clinicopathological characteristics at primary melanoma diagnosis for cutaneous melanoma patients with/without brain metastasis were selected for meta-analysis. Risk factors found on meta-analysis were then assessed in multivariate analyses using data from two, prospectively-collected melanoma databases at New York University (NYU) Medical Center. Receiver operating characteristic curves were used to test the predictive utility of these risk factors and to determine the optimal cut-point differentiating low- vs. high-risk primary melanoma patients. Melanoma brain metastasis-risk prediction models built using primary melanoma characteristics can be used to discriminate patients with/without brain metastasis and may be used to target high-risk patients for surveillance/prevention trials.
Other Project descriptions will be updated soon.
