A data-driven approach to identify protective drugs in COVID-19

A study performed in Ticino between Ente Ospedaliero Cantonale (EOC), USI Università della Svizzera italiana and Università Vita-Salute San Raffaele (UniSR, Milan, Italy) has shown how drugs against hypertension can reduce by more than 60% the risk of mortality in COVID-19 patients. The multidisciplinary study concerned 576 patients admitted to the EOC during the first wave of the epidemic, and was published in the prestigious journal Proceedings of the National Academy of Sciences (PNAS).

The study, based on data from 576 patients admitted to the EOC between March 1 and May 1, 2020 with an average age of 72 years, brought together a multidisciplinary team of clinicians and statistical researchers from the EOC, USI, and UniSR. The team work has revealed, through advanced statistical analysis of patient demographic and clinical data integration, that common anti-hypertensive therapies with renin-angiotensin system inhibitors - the so-called "RAASi" drugs - reduce by more than 60% the risk of mortality in COVID-19 patients considered at higher risk of death because they are older and/or have renal and cardiovascular diseases.

For the first time using a sophisticated statistical approach, researchers derived different risk profiles to assess the effect of drugs, analyse dependencies between different risk factors, and the impact of treatments on survival. The observed effect of RAASi is likely to be attributed to the interaction between the coronavirus and the renin-angiotensin system itself. In fact, it is known that SARS-CoV-2 enters host cells after binding to the angiotensin-converting enzyme (ACE2), whose function it "blocks", causing an excess of angiotensin and an increase in inflammation in the body, inflammation that is reduced by RAASi drugs.

Clelia Di Serio, Full professor of Medical Statistics at UniSR and Adjunct professor at USI explains:

"The main difficulty in analysing the data of this pandemic lies in the nature of the collection on an emergency basis, so statistical-computational techniques are needed to balance numerically unbalanced risk groups and to consider confounding effects.

With researchers Federica Cugnata and Chiara Brombin, we applied a combination of non-parametric and machine learning techniques to derive the complex dependency structure between treatments, comorbidities, risk factors and clinical response. The interaction between all researchers involved allowed a thorough reading of the models, based on reproducibility principles, and confirmed generalizable results: we consider the result on RAASi protection very solid."