Despite much progress, translational research development in the clinic is still lagging far behind in most areas of medical and clinical research. Drug development and clinical trials are still expensive for all stakeholders, with a very high failure rate. More than two-thirds of the compounds die in a stage known as the "Death Valley". It would cost $1.5 billion to $2.5 billion for a drug to be successful – moving fully through all stages of drug development to the clinic. This, combined with the inherent inefficiencies and inadequacies that plague the healthcare system, has led to a crisis in clinical research. Therefore, innovative strategies are needed to engage patients and generate the necessary evidence to drive new clinical advances to improve public health.
The process of drug discovery and development, and the failure rate at each step.
The average total cost of developing a marketed drug is 17500 million US dollars, 90% of clinical drug candidates will fail in the research and development process, and the research and development of innovative drugs is full of unknown challenges, like throwing dice at God. Especially in recent years, relying on the negative results of the phase II single-arm rapid approval of the drug indication, the negative results in the phase III trial have been tragically withdrawn. Like the withdrawal of academic **, the R&D of the industry has re-examined some positive signals of Phase II, can a positive result of Phase III be guaranteed?The point of contention is certainly not the correlation between ORR and OS, but the predictability of PFS on the outcome of OS. As a result, the research and development of new drugs, compared with the past, has added more uncertainties to the immune track represented by immune checkpoint inhibitors.
Prior to the pandemic, 30 years of clinical research had remained almost unchanged, and some trial codes of conduct and rules, albeit archaic, were unquestionable. The pandemic has disrupted, catalyzed, and accelerated innovation in the field by exposing many of the inherent systemic constraints in the trial process and forcing clinical trial research companies to re-evaluate all processes. Lessons learned should help researchers design and implement the next generation of "patient-centric" clinical trials.
The subspecialization of clinical specialties leads to divisions within and between specialties;Each major disease area appears to operate entirely independently. However, the best clinical** is delivered in a multidisciplinary manner, where all relevant information is available. Better clinical research should leverage the knowledge gained from each specialty and achieve a collaborative model that leads to multidisciplinary, high-quality continuous innovation in medicine. Although a single model may not be applicable for all diseases, interdisciplinary collaboration will allow the system to generate the best evidence more efficiently.
Over the next decade,Applications of machine Xi, deep neural networks, and multimodal biomedical AIClinical research will be reinvigorated from all angles, including drug discovery, image interpretation, streamlining electronic health records, improving workflows, and contributing to the optimization of public health systems over time (Figure 1). In addition, innovations in wearables, sensor technology, and Internet of Medical Things (IOMT) architectures offer many opportunities and challenges for capturing data. From this perspective, Prof. Vivek Subbiah shared his vision for the future of clinical trials and evidence generation, and identified areas for improvement in the areas of clinical trial design, clinical trial conduct, and evidence generation.
Timeline for drug development from now to the future.
This number represents the timeline from drug discovery to first-in-human Phase 1 clinical trials and finally FDA approval. Phase IV clinical trials are conducted after FDA approval and may last for several years. Over time, there is an urgent need to optimize clinical trial processes through drug discovery, interpretive imaging, streamlining electronic health records, and improving workflows. At all stages of drug development, AI can help in many ways. Deep Neural Networks;Electronic Health Record;IOMT, Internet of Medical Things;Machine learning.
-- Clinical study design --
Trial design is one of the most important steps in clinical research. Better protocol design leads to better clinical trial progress and faster "pass-or-no-go" decisions. In addition, the losses caused by poor design and failed experiments are not only economic losses, but also social losses.
Challenges of randomized controlled trials
Randomized controlled trials (RCTs) have been the standard for evidence generation in all areas of medicine, as they allow for an estimate of the effect of a trial group in the absence of confounders. Ideally, every drug or intervention should be studied through well-controlled randomized controlled trials. However, conducting randomized controlled trials is not always feasible because of challenges in terms of timely evidence generation, cost, design of narrow populations, ethical barriers, and the time required to conduct these trials. By the time they were completed and published, randomized controlled trials quickly became obsolete (personal extension: in the case of the current boom in immuno-oncology**, many studies start with a control group designed with chemotherapy or a targeted** control group, and by the time they are completed, the standard regimen has been changed to an immunotherapy combination regimen). In addition, trial design is isolated, and many clinical questions remain unanswered. Therefore, the traditional experimental design paradigm must adapt to the contemporary rapid developments in genomics, immunology, and precision medicine.
Advances in clinical trial design
Clinical practice requires high-quality evidence, which has traditionally been achieved through randomised controlled trials. Over the past decade, substantial progress has been made in the design, implementation, and enforcement of the "master" protocol (the overarching protocol applicable to several sub-studies), which has led to a number of changes in clinical practice that have somewhat improved the stagnation of randomized controlled trials. Methodological innovation to address this need involves the joint evaluation of the effects of multiple** on multiple patients or multiple diseases within the framework of an overarching trial. The result of this innovation is calledParent programmaster protocol), is a holistic protocol designed to solve multiple problems. The parent protocol can be designed to evaluate one or more interventions in multiple diseases, or multiple interventions in a single disease (according to current disease classifications), each targeting a disease subtype or a specific population defined by biomarkers. Under this broad definition of the parent protocol, there are 4 different protocols:Umbrella studies, basket studies, platform studies, and master observational trials (MOTs).(Figure 2). All four are made up of a series of experiments or sub-institutes. These trials or sub-studies share key design and implementation elements and are therefore able to provide better coordination than when they were designed and implemented independently.
The category of the master agreement.
The master programme consists of four different categories of studies, namely:Basket research, umbrella research, platform researchwithPrimary observational test (MOT).。Each is a unique trial design that can include independent factors with control interventions and can be analysed individually or collectively, for added flexibility.
Further reading: Basket Research:
Basket trials were first proposed in 2014 by the American Association for Cancer Research (AACR) as an innovative clinical trial for precision cancer medicine. Basket assays refer to the targeting of multiple tumor types with the same gene mutation as the research object, rather than based on the tumor site and pathological type, breaking the framework of tumor location and morphological characteristics. That is to say, as long as the patient has a tumor with a specific gene mutation, it can be considered to be included in the clinical trial for specific targeted drug data, which is essentially a drug to deal with different tumors. In short: cover a wide range of diseases or histological features (i.e., cancer). Once participants are screened, participants with a positive target are included in the trial, so the trial may involve many different diseases or histologic features. The parent protocol of a basket trial can contain many layers of different biomarker-drug combinations being tested.
Umbrella Study:
An umbrella study is a prospective clinical trial that evaluates multiple targets for a single disease stratified into multiple subgroups (based on sexual biomarkers or other patient risk factors). The umbrella test compares a certain type of disease to an umbrella, such as holding up a large umbrella (such as lung cancer), gathering different lung cancer driver genes, such as KRAS, EGFR, ALK, etc., under the same umbrella, which is to complete the detection of different targets at the same time, and then assign different precise targeted drugs according to different target genes. To put it simply: the test involves screening patients for a biomarker or other trait, and then assigning them to different tiers based on the results of the screening. Multiple drugs are studied in each layer, either by randomization design or by using an external control, depending on the disease.
Platform Research:
These are multi-arm, multi-stage study designs that compare several intervention groups with a common control group in the context of the same parent protocol definition. In addition, the series did not have a definite end date) and was more effective than the conventional trial due to the shared control group, which ensured that the proportion of patients in the intervention experimental group was higher than that of the control group. The COVID** (recovery) platform study is a prominent example;This practice-changing trial established dexamethasone as an effective method for COVID-19 and also showed that hydroxychloroquine was ineffective. Such trials can also be used for digital mental health interventions and can be easily implemented in resource-constrained settings.
Primary observational test (MOT).
the mot is a prospective, observational study design that broadly accepts patients independently of biomarker signature and collects comprehensive data on each participant. the mot is a combination of the master interventional trial and prospective observational trial designs and attempts to hybridize the power of biomarker-based master interventional protocols with the breadth of real-world data (rwd). this approach could be well suited to collect prospective rwd across many specialties; the registry of oncology outcomes associated with testing and treatment (root) mot is one example。(I'm afraid that the translation is inaccurate, but it is easier to understand in the original English text).
Patient-centered clinical research.
The main components of clinical research – patients, sites, sponsors (large and small pharmaceutical companies),* cooperative group sponsors, regulators, patient advocacy organizations, and CROs – need to work together to put patients at the center of clinical trials.
evidence-based deep medicine iceberg.
The current evidence-based medicine (EBM) pyramid is just the tip of the iceberg, providing little shallow evidence to care for the average patient. Therefore, there is a need for deep synthesis and fusion of all available data to enable the next generation of deep evidence-based medicine. The main challenge over the next two decades will be to extract, collate and mine the vast amounts of natural history data, genomics and all omics analyses, all published clinical studies, data collected by RWD and IOMT to provide next-generation evidence for deep medicine. pro, patient-reported outcomes.
References. subbiah v. the next generation of evidence-based medicine. nat med. 2023 jan;29(1):49-58. doi: 10.1038/s41591-022-02160-zif: 82.9 q1 . epub 2023 jan 16. pmid: 36646803if: 82.9 q1 .