drugs, often with weak methodology; even with meta-analysis, the information was not of sufficient quality to inform clinical practice. A comprehensive clinical trial system is required for a pandemic response, to ensure a sustained and effective pipeline of drug development across clinical trial phases. Clinical trial legislation requires that a trial is either open and recruiting or that it is not, and therefore it is currently logistically and financially challenging to maintain large clinical trial networks outside of pandemic settings. Some of the most nimble and responsive studies (International Severe Acute Respiratory and Emerging Infection Consortium and REMAP-CAP) have been those created to address epidemics and seasonal outbreaks that were consequently able to pivot rapidly to address the pandemic. It is anticipated that some of the platform studies will evolve to address common infections, such as influenza. While this will require a significant long-term resource commitment and legislative support, it offers the exciting possibility of rapidly generating definitive therapeutic guidance in a way not previously possible. The greatest advances have come from using existing tools better (e.g., data linkage, adaptive platforms) rather than inventing new ones. Interestingly, we have yet to see novel insights from pooling datasets.The most important innovation came with large, pragmatic outcome adaptive platform studies: RECOVERY, SOLIDARITY, PRINCIPLE, and ACTIV (2–5). The largest, RECOVERY, has provided practice-changing evidence of benefits in hospitalized patients for three therapies (dexamethasone, tocilizumab, and the neutralizing monoclonal antibody combination casirivimab/imdevimab) and ruled out significant benefits for six therapies (aspirin, azithromycin, colchicine, convalescent plasma, lopinavir/ritonavir, and hydroxychloroquine). Adaptive platform studies represent a step forward for demonstrating clinical effectiveness that previously was based on manufacturer-sponsored phase 3/4 studies. Their key features are a clear focus on hard endpoints with an economy of data collection, commonly using routinely collected data. They are adaptive, allowing ineffective interventions to be dropped and new candidates brought in.Rapid delivery of platform studies required an unprecedented level of cooperation between trial teams, drug manufacturers, regulators, and health system managers. This unified “top-down” approach is necessary to prioritize candidates for testing; large numbers are needed for definitive data, so the number of candidates that can be tested is limited. It has taken longer to formalize effective oversight processes than to start the studies themselves.Without question, the seemingly rapid development of COVID-19 vaccines has been a cornerstone of the COVID-19 response and might raise comparisons to the relative speed of drug development for COVID-19. Such comparisons highlight the reason for the rapid appearance of multiple COVID-19 vaccines: decades of investment in preparatory work, refining both science and application via strong collaborative efforts, leading to technology that was ready to be repurposed when needed. While the process of drug development is arguably subject to greater variability, the greatest successes in COVID-19 drug development so far have come where strong scientific investigation was in place many years prior to the emergence of the pathogen. Readiness for the next pandemic requires similar investment in preparatory work, with strong collaboration and sensible oversight.Outcome measures affect the speed of drug development. Phase 3 studies work best when fed preliminary data from fast, efficient phase 2 studies driven by robust surrogate markers for the clinical endpoint, namely death. The absence of surrogate markers in COVID-19 has meant that most phase 2 studies have employed clinical outcomes (World Health Organization [WHO] ordinal scale). Such tools require large numbers, leading to long recruitment times. As a result, data have not been available in a timely manner to inform drug selection for phase 3 platform trials. The phase 2 CATALYST study was one of very few to employ a Bayesian rather than frequentist model-based design using a biological surrogate (C-reactive protein, CRP) (6). This approach showed a clear difference between namilumab (antigranulocyte-macrophage colony-stimulating factor, anti–GM-CSF) and infliximab (antitumor necrosis factor, anti-TNF) based on small numbers of subjects (30 to 60 per arm), although neither agent has completed phase 3 testing to validate the predictive value of this surrogate.It is clear we need to establish reliable surrogates for death in order to expedite phase 2 trials. This will become increasingly more important as combination therapies become the norm and regulators insist that combination therapy includes treatments that are difficult to access, such as mAbs.The clinical course of COVID-19 is heterogeneous and may be prolonged. A major challenge with drug development has been tailoring therapies to the stage of disease. Most clinical trial data currently apply to hospitalized patients. There is limited insight into effective treatments for the early stages or in the posthospital or long-COVID setting.There is a window of missed opportunity: Early treatment, by reducing progression