APPLICATION OF ARTIFICIAL INTELLIGENCE FOR THE COMPARISON OF NEW DRUGS AND MEDICAL DEVICES
European Statement
Education and Research
Author(s)
Damuzzo V (1), Rivano M (2), Cancanelli L (3), Brunoro R (4), Gasperoni L (5), Ossato A (6), Colicchio A (7), Del Bono L (8), Di Spazio L (9), Celentano Fasano CN (10), Chiumente M (11), Mengato D (12), Messori A (13)
1) UOC Farmacia, AULSS2, P.O. di Vittorio Veneto
2) Hospital Pharmacy, Azienda Ospedaliero Universitaria, Cagliari
3) UOC Farmacia, AULSS2 Marca Trevigiana, P.O. di Castelfranco
4) School of Specialisation in Hospital Pharmacy, University of Milan, Milan
5) Oncological Pharmacy, IRCCS Istituto Romagnolo per lo Studio dei Tumori (IRST) ‘Dino Amadori, Meldola;
6) School of Specialisation in Hospital Pharmacy, University of Padua, Padua
7) Hospital Pharmacy, Azienda Unica Sanitaria Locale di Bologna (AUSL), Bologna
8) Azienda Ospedaliera Universitaria Pisana, Pisa
9) Hospital Pharmacy, Santa Chiara Hospital, Trento, Azienda Provinciale per i Servizi Sanitari (APSS);
10) Hospital Pharmacy, Azienda Ulss 3 Serenissima, Mirano,
11) Italian Society of Clinical Pharmacy and Therapeutics (SIFaCT), Milan
12) Hospial Pharmacy Unit – Azienda Ospedale-Università Padova
13) HTA Unit, Tuscany Region, Florence
Why was it done?
The clinical selection of available treatments and medical devices (MDs) is often hindered by the absence of direct efficacy comparisons between emerging therapies. This AI-tool aimed to address this challenge by employing advanced analytical techniques to facilitate informed decision-making in clinical settings.
What was done?
In 2016, the Italian Society for Clinical Pharmacy and Therapeutics (SIFaCT) launched the AVVICINARE project with the goal of training young hospital pharmacists to develop innovation in research based on non-original, already published data. We recently approached the field of indirect comparisons, applying the artificial intelligence (AI) technique ‘IPDfromKM’ to extract individual patient data (IPD) from Kaplan-Meier (KM) survival curves, enabling the indirect comparison of emerging pharmacological treatments and MDs
How was it done?
Drugs and technologies with similar therapeutic roles and efficacy assessed by time-dependent endpoints (Overall Survival, Progression-Free Survival) were identified. KM curves from relevant clinical trials were digitized, and the IPDfromKM application was used to reconstruct the IPD. Data from different studies on the same treatments were pooled to enhance sample size, and standard statistical techniques (Cox regression, inter-treatment comparison) were employed, considering long-surviving patients (restricted mean survival time [RMST]). A heterogeneity analysis ensured comparability of patient cohorts.
What has been achieved?
Drugs and technologies with similar therapeutic roles and efficacy assessed by time-dependent endpoints (Overall Survival, Progression-Free Survival) were identified. KM curves from relevant clinical trials were digitized, and the IPDfromKM application was used to reconstruct the IPD. Data from different studies on the same treatments were pooled to enhance sample size, and standard statistical techniques (Cox regression, inter-treatment comparison) were employed, considering long-surviving patients (restricted mean survival time [RMST]). A heterogeneity analysis ensured comparability of patient cohorts.
What next?
Given the increasing value of indirect comparisons in both clinical and pharmacoeconomic contexts, ongoing efforts will focus on refining the analytical techniques and expanding training programs for hospital pharmacists. Future work will also explore additional therapeutic areas to broaden the impact of evidence-based medicine and enhance the role of hospital pharmacists in clinical decision-making.