Explore chapters and articles related to this topic
Introduction
Published in Andrew P. Grieve, Hybrid Frequentist/Bayesian Power and Bayesian Power in Planning Clinical Trials, 2022
Part of the problem has to do with the basis on which the sample size in a study is chosen. I have argued previously (Grieve, 2015) that whilst it is possible to take a pragmatic decision about the sample size and to base it on what the program budget allows, this type of resource-sizing is unsatisfactory because it is associated with underpowering as Freiman et al. (1978) and Halpern et al. (2002) have shown. From a pharmaceutical perspective, it has been recognised over the last 20 years that the high failure rate particularly in late-phase clinical rates with average failure rates being as high as 45% (Kola and Landis, 2004) and as high as 60% in some therapeutic areas is unsustainable, although there are some signs of recent improvement (Hay et al., 2014). It can be argued that one element of the high failure is due to the tendency for development teams to be optimistic about the likely benefit of their drug candidate. This is understandable as by the time a candidate drug reaches the late stage of drug development, some members of the team may have spent a considerable proportion of their career on the development. My experience is that there are overt, and covert, incentives for teams to be optimistic and a negative consequence of such enthusiasm is likely to be an under-powering of studies. One approach to this problem is to get teams to be more realistic by acknowledging uncertainty in their view of the likely magnitude of the benefit of their development compound. This idea has been the subject of proposals for over 80 years.
AI/ML in Medical Research and Drug Development
Published in Wei Zhang, Fangrong Yan, Feng Chen, Shein-Chung Chow, Advanced Statistics in Regulatory Critical Clinical Initiatives, 2022
In the current highly competitive landscape of drug development, big pharmaceutical companies are utilizing AI-related methods more than any time before in different stages of drug development. Drug discovery and target identification, as one of the crucial stages of drug development and pharmaceutical research, is certainly not an exception. High failure rate of clinical trials [Mullard, 2018] as well as enormous costs of running them are the main contributors. As a result, exploiting ML based approaches can help making the process more efficient and cost effective. As suggested by the title of this section, the application of AI/ML methods in this stage of drug development can be divided into two parts: target identification and molecular design.
Bayesian Statistical Methodology in the Medical Device Industry
Published in Emmanuel Lesaffre, Gianluca Baio, Bruno Boulanger, Bayesian Methods in Pharmaceutical Research, 2020
In a clinical trial, the primary focus is on clinical outcomes, such as the failure rate in patients, θ. That is, for real patient outcomes y, we assume a standard distribution f(y|θ) indexed by the parameter of interest θ. If the SEM accurately predicts clinical outcomes, then the outcomes of virtual and clinical patients will be similar. Therefore, the virtual patients should provide relevant information about θ. Using this assumption, we interpret y0 as having been generated from a distribution governed by the clinical parameter θ (i.e. Y0 ∼ f(y0|θ) instead of generated from p0 (y0| D0) — Equation (25.1), which, as mentioned above, may be complex with no closed form. This interpretation is done for notational and mathematical convenience, as the power prior approach discussed in detail in the beginning of Section 25.3.3 requires the selection of a closed form likelihood function governed by the clinical parameter θ. This assumption is similar to standard statistical practice, where a standard distribution (e.g. Normal) is selected because it fits the data reasonably well. Data from a well-developed SEM will frequently be highly similar to clinical data.
Hormonal and natural contraceptives: a review on efficacy and risks of different methods for an informed choice
Published in Gynecological Endocrinology, 2023
Andrea R. Genazzani, Tiziana Fidecicchi, Domenico Arduini, Andrea Giannini, Tommaso Simoncini
In general, many factors can affect the failure rate of a contraceptive method, and this is true also for HCs. According to a study by Bradley et al. age was the main responsible for changes in contraceptive failure and adolescents consistently experienced the highest failure rate. Moreover, failure rates, particularly for oral contraceptive pills, were substantially higher for women in the poorest quintile of population compared to those in the wealthiest households. Failure was generally more relevant for users of short-acting HCs compared to long-acting HCs. However, HCs were the safest contraceptive methods, with users of condom, withdrawal, or periodic abstinence experiencing the highest failure rates [10]. A regular lifestyle and a stable sexual relationship favor the efficacy of contraceptive methods, particularly FABM, since they are related to the knowledge of one’s own fertility and to the regular monitoring of one’s own secretions and body changes. In addition, a regular menstrual cycle can help a woman to identify the fertile and non-fertile days of the month. Adolescents may be less likely to use contraception regularly for many reasons: they may have irregular menstrual cycles, they may experience more changes in their daily lives, they may have less stable relationships, and they may have occasional sexual relationships. All of these aspects can affect compliance with the chosen method of contraception. Therefore, a contraceptive method that is less dependent on daily compliance, such as an IUD, vaginal ring, or transdermal patch, may be the best choice for them.
A systematic review of invasive pneumococcal disease vaccine failures and breakthrough with higher-valency pneumococcal conjugate vaccines in children
Published in Expert Review of Vaccines, 2022
Bruce A. Mungall, Bernard Hoet, Javier Nieto Guevara, Lamine Soumahoro
Thirteen studies enabled the calculation of the overall vaccine failure rate. Across these studies, there were 429 vaccine failures out of a total of 5,114 IPD cases in vaccinated children, resulting in a failure rate of 8.4%. Ten studies enabled the calculation of the overall breakthrough IPD rate, which was 9.3% (361/3,888 across studies). Rates varied substantially according to study and vaccination schedule. Average vaccine failure rates were 19.0% (267/1,403) across two studies assessing a 3+0 schedule (both in Australia [28,36]), 4.2% (130/3,097) across six studies assessing a 2+1 schedule (in France [23], United Kingdom [25], Switzerland [27], South Africa [29], Canada [6] and Morocco [31]) and 5.2% (32/614) across five studies evaluating a 3+1 schedule (in Portugal [24], Spain [32,34,37] and Germany [46]). Differences in study settings, populations and follow-up periods may have, however, influenced these results.
Self-Defeating Codes of Medical Ethics and How to Fix Them: Failures in COVID-19 Response and Beyond
Published in The American Journal of Bioethics, 2021
Despite the fact that current approaches to drug development leverage large data sets to generate hypotheses of the sort described above to identify drug candidates for development, roughly 90% of such interventions are never approved for any indication (Hay et al. 2014; Thomas et al. 2016; Wong et al. 2019). During this process, it is common for interventions to show significant promise in small, early phase trials but to fail in large, confirmatory trials (also known as phase 3 trials). Moreover, this failure rate is somewhat generous since drugs that enter the pipeline as candidates to treat one condition sometimes end up being approved for a different indication. While that might be a great boon to investors, it isn’t much consolation for the patients with the condition that was the initial target for development. Likewise, regulatory approval is not always a reliable proxy for improving clinical outcomes that matter to patients since many interventions are approved on the basis of surrogate endpoints that may not track clinically meaningful gains in survival (Kemp and Prasad 2017).