EPICLIN 2025 / JSCLCC 2025
As part of this annual gathering of the French-speaking community in clinical epidemiology and biostatistics applied to biomedical research, numerous topics were addressed, including methodological innovations for clinical trials, the contributions of artificial intelligence to clinical research, and approaches in causal inference.
The PEPITES and CLEPEVIR teams from IPLESP were present, with involvement starting at the organizational level thanks to Prof. Florence Tubach (PEPITES), a member of the scientific committee. Three presentations from our institute were featured.
Bertrand Bouvarel (CLEPEVIR) won the award for Best Oral Presentation for his work entitled “Causal inference using synthetic data generated from electronic health records as an alternative to inverse probability weighting: a simulation study.” This exploratory study compared the use of different generative models (GANs, transformers, CART) to a standard IPTW approach for emulating randomized trials. The results highlight the potential of synthetic data—especially with transformers and CART—offering comparable performance to standard IPTW while reducing the impact of extreme weights and increasing the effective sample size.
Maxime Beydon (PEPITES) delivered an oral presentation on the emulation of a clinical trial using the French national health data system (SNDS), assessing the benefit-risk balance of aspirin in the primary prevention of cardiovascular disease at the time of diagnosis of giant cell arteritis. The methodological approach was based on cloning, censoring, and weighting, and included a negative control outcome and exposure to evaluate residual confounding. Results suggest a balance between cardiovascular benefits and bleeding risks at one year, with a sex-specific effect: a benefit in women, but not in men.
Cyrille Guillot-Tantay (PEPITES) also presented a trial emulation using SNDS data in a poster session, focusing on the use of mid-urethral slings in the management of stress urinary incontinence. Long-term complications of the two main types of slings—retropubic and transobturator—were compared using a Cox model weighted by propensity scores. The risk of sling removal or cutting was higher with TVT slings than with TOT slings, with a time-dependent effect.
The conference also offered high-quality scientific exchanges under the guidance of an organization committed to promoting environmentally responsible scientific events.