Le choix des critères de jugement (principal et secondaires) est une étape essentielle de la construction d’un essai contrôlé randomisé. Notre premier travail a été de réaliser une revue systématique sur la place des critères de jugement importants pour les patients ou patient-important outcomes dans les essais contrôlés randomisés en réanimation. Nous avons défini les patient-imortant outcomes comme étant la mortalité d’une part, et l’impact fonctionnel et sur la qualité de vie des séquelles de la réanimation d’autre part. En effet, en réanimation, les deux objectifs thérapeutiques principaux sont d’une part éviter le décès des patients et d’autre part réduire l’impact des séquelles à moyen et long terme chez les survivants. Nous avons montré que les patient-imortant outcomes ne représentent qu’une proportion faible (27/112, 24%) des critères de jugement principaux des essais publiés en 2013 et que pour une grande majorité il s’agissait d’un critère de mortalité. Une analyse sur les essais publiés au premier semestre 2016 a montré qu’il n’y avait pas eu d’évolution (25% des critères de jugement principaux étaient des patient-important outcomes).Puis, en réalisant une étude ancillaire de cette revue systématique, nous avons investigué l’impact potentiel des décisions de limitation(s) ou d’arrêt(s) des thérapeutiques actives quand le critère de jugement était la mortalité, et décrit les données rapportées sur ces limitation(s) ou d’arrêt(s) des thérapeutiques actives dans les essais en réanimation. Nous avons montré que très peu d’essais contrôlés randomisés en réanimation (6/65, 9%) rapportaient le taux de décision de limitation(s)ou d’arrêt(s) des thérapeutiques actives, bien que ces décisions soient fréquentes en pratique clinique. Pour explorer l’impact qu’un déséquilibre de ces décisions entre les 2 bras d’un essai pouvait avoir en termes de biais sur la mortalité, nous avons réalisé une étude de simulation. Cette étude a montré notamment que lorsque ces décisions étaient prises de façon plus tardive dans le groupe expérimental, l’intervention pouvait apparaître protectrice alors même qu’il n’existait pas de réel effet sur la survie. Enfin, nous avons conduit un essai contrôlé randomisé en réanimation (étude AKIKI, Artificial Kidney Initiation in Kidney Injury) en utilisant la mortalité comme critère de jugement principal et en rapportant le taux et le délai des décisions de limitation(s) ou d’arrêt(s) des thérapeutiques actives dans les 2 bras / The choice of relevant primary and secondary endpoints is an essential step of the design of a randomized controlled trial. In our first work, we conducted a systematic review on patient-important outcomes in randomized controlled trials in critically ill patients. Indeed, clinical decision-making by ICU physicians now pursues the goal of improving mean and long-term outcomes in survivors in addition to increasing their chance of survival. We defined patient-important outcomes as on one hand, outcomes involving mortality at any time, and on the other, quality of life and functional outcomes assessed after ICU discharge. We found that a minority of primary outcomes (27/112,24%) used in randomized controlled trials published in 2013, were patient-important outcomes and that mortality accounted for the vast majority of them. Our analysis of most recently published trials (first half 2016) showed that patient-important outcomes were used in the samelow proportions (25% of the primary outcomes were patient-important outcomes) We then addressed the question of how well withholding and with drawal of life support therapies(W-WLST) decisions were reported in RCT in critically ill patients and how such decisions could impact mortality as outcome measure in these trials. We found that W-WLST decisions, although being a daily concern in routine practice, were scarcely reported in these trials, since they appeared in only 6 of 65 (9%) during follow-up. We further explored the impact of an imbalance in such decisions between the 2 arms of a randomized controlled trial, through a simulation study. This simulation showed that the intervention could appear as protective, if the decision of W-WLST was delayed in the interventional arm, even though the intervention had no true effecton survival. Finally, we performed a randomized controlled study (Artificial Kidney Initiation in Kidney Injury,AKIKI) using mortality as primary outcome and paid attention to report the rate of W-WLSTdecisions in the 2 arms.

Background Opioid use has become a huge public health crisis and opioids are now one of the leading causes of deaths related to drugs worldwide. Identifying differences in predictors and treatment outcomes for people with opioid use disorder (OUD) that were introduced by prescription versus other means is important. It is also vital to understand what the goals are and needs that patients want to achieve out of OUD treatment. Methods We used systematic review methodology to first examine any adverse outcomes that may be associated with prescribing opioids for acute low back pain. We also conducted a systematic review and meta-analysis examining what are the differences in patients with OUD that were initially introduced to opioids by prescription in comparison to those introduced by recreational means. We then conducted an observational study using data obtained from the GENetics of Opioid Addiction (GENOA) research collaborative. We examined treatment outcome differences between individuals introduced to opioids through a licit prescription and those introduced through illicit means. We conducted a mixed-methods study asking what the desired goals of patients with OUD from the Pharmacogenetics of Opioid Substitution Treatment (POST) project are. Using data from POST, we also examined the treatment outcome differences between those that were receiving methadone treatment in comparison to those that were on buprenorphine. Results The systematic review examining adverse outcomes of prescribing opioids for acute low back pain found that prescribing opioids for ALBP was significantly associated with long-term continued opioid use (1.57, 95% CI 1.06,2.33). The second systematic review found that those who were introduced to opioids through a legitimate prescription were significantly less likely to have illicit opioid use (0.70, 95% CI 0.50, 0.99) while in treatment. Our results from GENOA also showed that those introduced to opioids by prescription were more likely to have chronic pain, an older age of onset of opioid use, less likely to have hepatitis C and use cannabis. When we asked patients what goals they desired out of treatment, we found that the most frequently reported patient important outcomes were to stop treatment (39%) and avoid all drugs (25%). When comparing OUD patients by treatment we discovered that those receiving buprenorphine were less likely to consume illicit opioids and amphetamines but more likely to have used alcohol in comparison to those on methadone. Conclusion With this knowledge, we can recognize unique risk factors for each patient and provide more tailored treatment that can incorporate this into clinical practice to address specific concerns in various cohorts of OUD patients. Additionally, the variation in the selection of outcomes demand the need for further research to establish a set of outcomes that considers patients’ goals and preferences for OUD treatment. / Thesis / Candidate in Philosophy