clinical trials. These DOE should be explored to better design clinical trials. Objectives:To identify effect-modifiers of antipsychotic drugs, in schizophrenia treatment.Methods: Data were retrieved from the CGS observational cohort study, including 1859 schizophrenia inpatients and outpatients aged 15-65 years old through 177 centres in France. Patients were followed-up at 3, 6, 9 and 12 months.Patients who initiated or switched APD were identified and schizophrenia symptoms evolution was measured 3 to 6 months later, using the BPRS-18 scale (∆BPRS).First, potential DOE of APD were identified through a focused literature search, which was reviewed by 3 schizophrenia specialized psychiatrists. Five DOE were short-listed: (1) shorter duration of illness, (2) higher level of negative symptoms, (3) poor adherence, (4) cannabis/drug use and (5) tobacco use.Effect modification was assessed using sub-group analyses, with comparisons of the ∆BPRS in the 2 strata of each DOE. Two-sided Welch t-tests were used with a "non-conservative" type-I error α = 0.2. Multivariate analyses were not performed due to limited sample size.Results: Out of 1859 schizophrenia patients, 116 patients initiated drug B, 272 patients initiated drug D and 204 patients initiated drug K. Other drugs were initiated by too few patients. The mean decreases in BPRS were: -7.2 points (SD = 16.3) in "drug B initiators", -7.5 points (SD = 15.3) in "drug D initiators" -3.9 points (SD = 14.1) in "drug K initiators". The level of symptoms improvement was higher in patients with a "higher level of negative symptoms" for all drugs (p<0.012) and "poorer adherence", for drugs D and K (p<0.013). The level of symptoms improvement was also better in patients who did not smoke, however not significantly. Cannabis use was not an effect-modifier, in all drugs. Conclusions:Overall, "adherence", "tobacco smoking" and "negative symptoms" may be drivers of effectiveness. These factors should be adequately captured and explored in pre-launch clinical trials to avoid an efficacy-effectiveness gap. Background: Beneficial effects of drugs can be divided into efficacy and effectiveness. An understanding of the efficacy-effectiveness gap is important for patients, health care professionals, payers, regulators and the pharmaceutical industry to provide effective treatments. A Systematic Literature Review on the Efficacy-Effectiveness Gap: Comparison of Randomized Controlled Trials and Observational Studies of Glucose-Lowering Drugs

household members during baseline to determine prevalent household use. Results:We identified 218,594 prescription opioid and 114,767 prescription NSAID initiators between 2000-2013. Of these initiators, 160,404 (73% opioid) and 82,359 (72% NSAID), respectively, were enrolled within a household. Household size (median: 3 members, interquartile range (IQR): 2-4) and age of household members (median: 21 years, IQR: 12-45) were similar in both treatment groups. Prevalent household use of prescription opioids was 27% (95% confidence interval (CI): 27-28) among opioid initiators and 24% (95% CI: 24-24) among NSAID initiators. Prevalent household use of prescription NSAIDs was 20% (95% CI: 19-20) among NSAID initiators and 18% (95% CI: 17-18) among opioid initiators. Conclusions:In our new user cohort, we observed high levels of prevalent household opioid use among both treatment groups. Comparative safety studies should consider the household availability of medications as a potential source of exposure misclassification which may introduce prevalent-user bias into new-user designs. Background: Channelling bias may occur when a newly marketed drug and an established drug, despite similar therapeutic indications, are prescribed to patients with different prognostic characteristics (also known as confounding). Assessment of Channelling Bias Among Initiators of Glucose Lowering Drugs -A Cohort StudyObjectives: To investigate channelling bias and its impact on estimated effectiveness of GLP1 and DPP4, compared to basal insulin and sulfonylurea (SU), respectively. Methods:We used CPRD data to include patients with T2D, initiating treatment (6 months drug-free) between 2007 (when GLP1 and DPP4 were approved by EMA) and 2015. All analyses were split in 7 time blocks of 365 days each. The characteristics of patients initiating GLP1 versus basal insulin, and DPP4 versus SU were compared over time. After propensity score matching (based on sex, age, BMI, diabetes duration, Charlson comorbidity score, use of other glucose-lowering drugs, use of anti-hypertensives, statins, anti-coagulants and diagnoses of hypertension, renal disease, myocardial infarction, and stroke), relative effectiveness on 6 month changes in HbA1C (%) was estimated. Results:In total, 8,398 GLP1, 14,807 insulin, 24,481 DPP4, and 33,505 SU initiators were identified. Time trends for GLP1 and DPP4 showed that use of anti-hypertensives decreased among GLP1 initiators, and increased among DPP4 initiators. Use of other oral glucose-lowering drugs decreased, BMI decreased, and use of statins increased among DPP4 initiators. For other characteristics the difference between comparison groups were small and did not indicate channelling.Propensity score matched analyses included 4,072 pairs of GLP1 and insulin initiators, and 10,620 pairs of DPP4 and SU initiators. Conclusions:Channelling was not widespread and relative effectiveness appeared constant since launch of the drugs. It was possible to match many patients, which allowed for assessing relative effectivenes...