My research career has followed two main themes. In Houston, my lab focused on describing pharmacodynamic and immunological characteristics of antifungal agents used to treat mold opportunistic diseases in immunocompromised patients. Working with collaborators at the University of Texas MD Anderson Cancer Center, my laboratory clarified parameters associated with effective dosing for the antifungal echinocandins and amphotericin B used to treat invasive aspergillosis. More recently, we explored the pharmacokinetic-pharmacodynamic parameters associated with effective liposomal amphotericin B or posaconazole treatment of mucormycosis. Our laboratory was also the first group to detect the immunosuppressive metabolite, gliotoxin, in human and experimental models of invasive pulmonary aspergillosis. With contributors to MD. Anderson, we were able to demonstrate that gliotoxin impairs both innate and adaptive immune responses against the fungus, and blocks host angioneogenesis in tissue during invasion.\n\nIn Bologna, my research shifted more towards the development of clinical research tools for the diagnosis and monitoring of serious infections. In collaboration with researchers at the "Seràgnoli" Institute of Hematology in Bologna, and investigators in microbiology and radiology, we developed new risk models and explored the potential role of CT pulmonary angiography and angiographies to improve the diagnosis of invasive fungal infections.\n\nSince coming to Padova, I have continued developing risk models for managing Gram-negative MDR pathogens in critically ill patient populations. Part of these efforts also involve developing new antibiotic monitoring tools to facilitate the appropriate use and dosing of antimicrobials, tailoring approaches for PK-PD optimization of antibiotic administration, and establishing epidemiological databases that can be used to monitor resistance trends and evaluate the impact of our interventions. \n\nA list of my current projects can be found on my personal web page: https://russlewisid.com/project/

Bayesian Meta-Analysis of Diagnostic Test Data in Medicine

A correct medical diagnosis is crucial when dealing with disease or infections. In an ideal world, we would have a set of studies that would guide us on the best approach for a particular diagnostic problem. Unfortunately, this is not always the case. Researchers often have to work with a fragmented and varied body of evidence that needs to be analyzed.

To tackle this issue, researchers use meta-analysis, a statistical technique that enables them to combine evidence from multiple sources. However, Diagnostic test data have unique challenges. Diagnostic summaries, such as sensitivity and specificity, can be interdependent, making a simple combination misleading. Moreover, diagnostic studies are typically conducted in different scenarios and populations, leading to high heterogeneity between study results. Lastly, the studies included may be of different qualities and designs, further complicating the meta-analysis process.

Recently Bayesian approaches have been developed to perform meta-analysis of diagnostic test data that allow for more accurate summary of diagnostic test performance by calculating the marginal and joint posterior predictive distribution of the sensitivity and specificity

For this thesis, the student is invited to select a diagnostic test used in medicine/infectious diseases where there is are unresolved clinical questions in the test performance and optimal clinical applications.

1. A detailed literature search will be performed to identify relevant studies which will be analysed and summarised in a database according to PRISMA-Diagnostic Test Accuracy Guidelines.

2. Bayesian Meta-Analysis models will be fit to the student's database under the supervision of Prof. Lewis using the BAMDIT package in R statistical computing language to estimate pooled sensitivity and specificity with predictive posteriors (measures of uncertainty) and identify outlier studies that should not be given weight for diagnostic assessment

3. Results will be used to describe generate heterogeneity of the studies included in the meta-analysis and the prediction of future diagnostic study results.

4. Recommendations for future studies or correct clinical application of the test will be provided.

Students considering the thesis area should contact Professor Lewis by email (russelledward.lewis@unipd.it) to discuss the possible topic and timeline for completion.