EPICLIN 2020 / Revue d’Épidémiologie et de Santé Publique 68 (2020) S97–S118 S113 Conclusions En présence d’interaction avec un modificateur dis- tribuée différemment dans les arrêtes, la méthode DA peut conduire à un bais dans l’estimation de l’ET. La méthode utilisant les DIP en une étape devrait être considérée comme celle de référence lors de l’analyse d’une méta-analyse en réseau. Mots clés Interaction ; Méta-analyse ; Réseau ; Données individuelles ; Données agrégées Déclaration de liens d’intérêts Les auteurs n’ont pas précisé leurs éventuels liens d’intérêts. https://doi.org/10.1016/j.respe.2020.03.007 9.2 Association between initial and residual pulmonary vascular obstruction and pulmonary embolism recurrence, a pooled analysis R. Chaux a,∗ , O. Sanchez b,c , F. Couturaud b,d , N. Meneveau e , R. Chopard e , P. Mismetti a,b,f,h , S. Laporte a,b,g , E. Ollier a,g a CHU de Saint-Étienne, unité de recherche clinique, innovation et pharmacologie, Saint-Étienne, France b F-CRIN, INNOVTE, Saint-Étienne, France c AP–HP, hôpital européen Georges-Pompidou, université de Paris, service de pneumologie et de soins intensifs, Inserm UMRS 1140, Paris, France d CHU de Brest, université de Bretagne Occidentale, département de médecine interne et pneumologie, EA 3878, CIC Inserm 1412, Brest, France e CHU de Besanc¸ on, département de cardiologie, EA3920, Besanc¸ on, France f CHU de Saint-Étienne, Inserm, CIC-1408, Saint-Étienne, France g Université de Saint-Étienne, Inserm, U1059, dysfonction vasculaire et hémostase, Saint-Étienne, France h CHU de Saint-Étienne, service de médecine vasculaire et thérapeutique, Saint-Étienne, France ∗ Corresponding author. Adresse e-mail : robinvchaux@gmail.com (R. Chaux) Introduction The best treatment option (short-term or exten- ded anticoagulant treatment) after a first unprovoked pulmonary embolism (PE) is debated. Pulmonary vascular obstruction, mea- sured by the pulmonary vascular obstruction index (PVOI) at diagnosis or during the follow-up, has been associated with an increased risk of venous thromboembolism (VTE) recurrence when anticoagulants are stopped [1]. But results are not entirely consistent [2], and PVOI classification cut-offs vary among studies. We therefore conducted a pooled analysis of 3 studies of the F-CRIN INNOVTE network. The primary objective of this work is to assess if elevated PVOI at the time of PE diagnosis and after months of anti- coagulation are independent risk factors of VTE recurrence after a first episode of PE, and if so, to estimate which PVOI cut-off values would be the most discriminating for predicting VTE recurrence. Method We performed a pooled analysis of individual data from 3 French studies conducted within the F-CRIN INNOVTE net- work: a randomised double-blind controlled clinical trial and two observational longitudinal cohort studies. PVOI was measured by ventilation/perfusion lung scan during two periods; one at the time of index PE (initial PVOI), and one several months later, after anti- coagulation treatment of the acute phase (post-treatment PVOI). PVOI was expressed in percentage of vascular obstruction. VTE recurrence was defined as objectively confirmed PE episode or deep vein thrombosis. The associations between PVOI (both ini- tial and post-treatment) and VTE recurrence were jointly estimated using a multivariate Cox proportional hazards regression model. To manage confounding, the model was further stratified on the study, as well as adjusted on anticoagulant treatment duration, which could have been affected by PVOI measurement in the two obser- vational cohort studies. By this means, the direct effect of PVOI on VTE recurrence may be estimated. Optimal PVOI thresholds were estimated by 10-fold cross-validation, using areas under curve (AUC) computed from time-dependant receiver-operating charac- teristic (ROC) curves. Results The total number of patients was 922. The median dura- tion of follow-up was 42 months (IQR: 38–60). Mean age was 60 ± 18 years, 56% of patients were female. VTE recurrences occur- red in 149 patients (4.2 events per 100 person-years). The most discriminating cut-offs were > 35% for initial PVOI (AUC: 0.71, 95% CI: 0.71–0.71), and > 5% for post-treatment PVOI (AUC: 0.69, 95% CI: 0.68–0.69). Initial PVOI > 35% and post-treatment PVOI > 5% were significantly associated with the risk of VTE recurrences (HR 1.61, 95% CI: 1.07–2.43 and HR 1.63, 95% CI: 1.06–2.50, respectively). Conclusion After a first episode of PE, initial PVOI > 35% and post- treatment PVOI > 5% were found to be independent predictors of VTE recurrence. Those results could in the future help detect more accurately patients with high-risk of VTE recurrence, and thus pro- vide help in clinical decision-making for adapting anticoagulation therapy. Keywords Pooled analysis; Pulmonary embolism; Veinous thromboembolism; Pulmonary vascular obstruction index; Cross-validation Disclosure of interest The authors have not supplied their decla- ration of competing interest. Références [1] Meneveau N, et al. Long-term prognostic value of residual pulmonary vascular obstruction at discharge in patients with intermediate-to high-risk pulmonary embolism. Eur Heart J 2013;34:693–701. [2] den Exter PL, et al. Thromboembolic resolution assessed by CT pulmonary angiography after treatment for acute pulmonary embolism. Thromb Haemost 2015;114:26–34. https://doi.org/10.1016/j.respe.2020.03.008 9.3 Estimating the difference in restricted mean survival time accounting for trial effect in individual patient data meta-analyses S. Jonas a,c,∗ , D. Zucker b , S. Michiels a,c a Institut Gustave-Roussy, Service de biostatistique et d’épidémiologie, Villejuif, France b The Hebrew University of Jerusalem, Department of Statistics and Data Science, Jerusalem, Israel c Inserm, CESP Inserm Oncostat, Villejuif, France ∗ Corresponding author. Adresse e-mail : sarah.jonas@gustaveroussy.fr (S. Jonas) Introduction The difference in restricted mean survival times (RMSTs) between two treatment groups is a useful tool to provide information on the average causal treatment effect in a randomised clinical trial. This method is particularly appealing since it does not require a proportional hazards assumption, unlike the hazard ratio based on the Cox model. The RMST can be obtained by integrating under the survival curve up to a predetermined horizon. A parti- cular concern in individual patient data meta-analyses (IPD-MA) is to properly account for the trial effect in the estimation of the difference in RMSTs. Objective Our objective is to estimate the difference in RMSTs in an IPD-MA using various propositions where the RMST is adjusted on covariates and includes a random trial effect. Methods In our first proposition, we use the Breslow estimator as proposed by Zucker (JASA, 1998) to have an estimation of the baseline hazard adjusted on the trials and on other covariates, and calculate the area under the curve. A variance estimator is obtained with an added term in the expression (Chen and Tsiatis, Biome- trics, 2001) to remove the conditioning on the trial. The second