What is known and objective: The real-life prognostic impact on long-term survival of continuous or discontinuous adherence to ESC guideline-recommended drugs in heart failure with reduced ejection fraction (HFrEF) patients has rarely been investigated. Here, we present the long-term association of longitudinal prescription of guideline-recommended drugs with 3-year all-cause and cardiovascular (CV) mortality in HFrEF patients. Methods: We used data from the EPICAL2 cohort study of 624 hospitalized HFrEF patients. Using the sequence analysis, we classified patients into five groups of long-term adherence according to the continuity/discontinuity of their prescription adherence to guidelines over a 3-year follow-up, as follow: 316 (50.6%) patients in the sustained adherence group, 163 (26.1%) in the sustained non-adherence group, 79 (12.6%) in the adherence to non-adherence group, 43 (6.9%) in the non-adherence to adherence group and 23 (3.7%) in the multiple switches group. The associations between all-cause mortality and CV mortality and the adherence groups were determined by Cox and Fine-Gray models, respectively. To account for immortal time bias, we performed a landmark analysis at 24 months. Patients who died, prior to the landmark time, were excluded from this analysis and long-term adherence groups were redefined. Results and discussion: After adjustment for confounding factors, as compared to the sustained non-adherence group, the sustained adherence group showed lower all-cause and CV mortality (hazard ratio HR = 0.37 [0.25-0.56] and sub-distribution hazard ratio SHR = 0.33 [0.20-0.56]). Both clinical outcomes were also significantly improved in the adherence to non-adherence group (HR = 0.25 [0.13-0.45] and SHR = 0.20 [0.10-0.41]), the non-adherence to adherence (HR = 0.24 [0.11-0.55] and SHR = 0.11 [0.04-0.30]), and for the multiple switches group (HR = 0.13 [0.07-0.51] and SHR = 0.12 [0.08-0.43]). Results from landmark analysis were comparable to the main results. What is new and conclusion: As in all observational studies, our results may be affected by residual confounding related to unmeasured confounders, although we attempted to adjust for many confounders. Even a discontinuous prescription of the recommended drugs over time was associated with better long-term outcomes. In other words, whatever the time of HFrEF evolution, prescribing recommended drugs at some point was always better than never prescribing.