Description
please read the document below
Unformatted Attachment Preview
Critical Analysis Assignment
Please read the following instructions before completing your critical analysis assignment.
The purpose of the critical analysis is to evaluate an article, book, paper, movie etc. and critique the written analysis
of the subject. It has a structure similar to an Essay writing and includes following two component s – first,
explain the author’s argument by understanding and analyzing the main findings/arguments in the article, and
second, provide your own argument by critically evaluating the article.
Structuring Critical Analysis
The critical analysis of an article should start with a critical reading by questioning the author’s argument on
the topic. It is important to be concise in all parts of your analysis and remain focused on your argument by avoiding
summary or irrelevant description. It should include an introduction, a summary of the work, analysis, and a conclusion.
The following outline can be used to write a critical analysis:
Introduction (1 Paragraph)
Provide the context/background information including the title of the article and the author. Briefly describe the
purpose or the main argument of the article. State your viewpoint about the article by providing the supporting
argument.
Summary of the Work (1-2 paragraphs)
Provide the main arguments or findings in the article by discussing the strengths, weaknesses and potential
solutions presented in the article.
Critical Analysis – your argument (1-2 paragraphs)
Explain what you liked or not liked about the author’s argument by providing specific examples from the article.
Conclusion (1 paragraph)
Summarize the main ideas from the critical analysis section and explain the importance of your argument. Explain the
potential for further research or analysis on your ideas and the overall importance of the article.
General Instructions:
•
•
•
•
•
Select one of the three articles to prepare your critical analysis document that includes an introduction,
summary of the work, critical analysis-your argument, and conclusion as described above.
o Article #1: Januszek et al. (2022) The role of management in lean implementation: evidence from the
pharmaceutical industry. International Journal of Operations & Production Management Vol. 43 No. 3,
2023 pp. 401-427.
o Article #2: Gould et al. (2022) A framework for assessing the impact of accelerated approval. PLoS ONE
17(6): e0265712. https://doi.org/10.1371/journal.pone.0265712
o Article #3: Vergara et al. (2023) Improving success rates by applying interventions in clinical practice and
measuring their impact: A multicenter retrospective analysis of more than 240,000 cycles. European
Journal of Obstetrics & Gynecology and Reproductive Biology. Volume 287, August 2023, Pages 186-194
The critical analysis should be between 500-1000 words.
Font: Times New Roman
Font size: 12
Please refer to rubric for more details on how the critical analysis will be evaluated
PLOS ONE
RESEARCH ARTICLE
A framework for assessing the impact of
accelerated approval
A. Lawrence Gould ID1*, Robert K. Campbell2, John W. Loewy3, Robert A. Beckman4,
Jyotirmoy Dey5, Anja Schiel6, Carl-Fredrik Burman7, Joey Zhou8, Zoran Antonijevic9, Eva
R. Miller10, Rui Tang11
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
1 Methodology Research, BARDS, Merck & Co., Inc., Kenilworth, New Jersey, United States of America,
2 Molecular Pharmacology, Physiology and Biotechnology, Brown University, Providence, Rhode Island,
United States of America, 3 DataForethought, Winchester, Massachusetts, United States of America,
4 Lombardi Comprehensive Cancer Center and Innovation Center for Biomedical Informatics, Georgetown
University Medical Center, Washington, DC, United States of America, 5 Data and Statistical Sciences,
AbbVie, North Chicago, Illinois, United States of America, 6 Department for Pharmacoeconomics, Norwegian
Medicines Agency, Oslo, Norway, 7 Data Science & AI, AstraZeneca R&D, Gothenburg, Sweden, 8 Xcovery
Pharmaceuticals, Palm Beach Gardens, Florida, United States of America, 9 Abond CRO, Allendale,
Michigan, United States of America, 10 Independent Biostatistical Consultant, Middletown Twp,
Pennsylvania, United States of America, 11 Methodology and Data Visualization, Biostatistics Department,
Servier Pharmaceuticals US, Boston, Massachusetts, United States of America
* [email protected]
Abstract
OPEN ACCESS
Citation: Gould AL, Campbell RK, Loewy JW,
Beckman RA, Dey J, Schiel A, et al. (2022) A
framework for assessing the impact of accelerated
approval. PLoS ONE 17(6): e0265712. https://doi.
org/10.1371/journal.pone.0265712
Editor: Ismaeel Yunusa, University of South
Carolina College of Pharmacy, UNITED STATES
Received: December 11, 2020
Accepted: March 7, 2022
Published: June 24, 2022
Copyright: © 2022 Gould et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
The FDA’s Accelerated Approval program (AA) is a regulatory program to expedite availability of products to treat serious or life-threatening illnesses that lack effective treatment alternatives. Ideally, all of the many stakeholders such as patients, physicians, regulators, and
health technology assessment [HTA] agencies that are affected by AA should benefit from
it. In practice, however, there is intense debate over whether evidence supporting AA is sufficient to meet the needs of the stakeholders who collectively bring an approved product into
routine clinical care. As AAs have become more common, it becomes essential to be able to
determine their impact objectively and reproducibly in a way that provides for consistent
evaluation of therapeutic decision alternatives. We describe the basic features of an
approach for evaluating AA impact that accommodates stakeholder-specific views about
potential benefits, risks, and costs. The approach is based on a formal decision-analytic
framework combining predictive distributions for therapeutic outcomes (efficacy and safety)
based on statistical models that incorporate findings from AA trials with stakeholder assessments of various actions that might be taken. The framework described here provides a
starting point for communicating the value of a treatment granted AA in the context of what
is important to various stakeholders.
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
information files.
Funding: The author(s) received no specific
funding for this work.
Introduction
Competing interests: The authors have declared
that no competing interests exist.
The FDA’s Accelerated Approval program (AA) is one of four FDA programs to expedite
availability of products that treat serious or life-threatening illnesses where effective treatments
PLOS ONE | https://doi.org/10.1371/journal.pone.0265712 June 24, 2022
1 / 19
PLOS ONE
Impact of accelerated approval
do not exist or where currently available treatments are inadequately effective or unacceptably
toxic (S1 Appendix). A product is a candidate for AA if it meaningfully affects a surrogate endpoint that is reasonably likely to predict clinical benefit or a clinical endpoint that can be measured earlier than irreversible morbidity or mortality (IMM) and is reasonably likely to predict
a subsequent effect on IMM [1].
From 1992 through mid-2021, the FDA granted 269 AAs for 162 different treatments
(drugs or drug combinations) [2]. Most were for infectious disease (24/162) and cancer (113/
162). The creation of the AA pathway was influenced by the HIV epidemic and patient advocacy. Thirty HIV treatments received accelerated approved using surrogate endpoints and all
fulfilled their post-approval study requirements [3]. The surrogate endpoints for efficacy
against HIV proved to be highly predictive of clinical outcome, and were eventually adopted
for traditional approval. These drugs and post-approval innovations in their use enable HIVinfected persons to have life expectancies close to the life expectancies of uninfected persons
[4]. There have been more than 110 AAs for malignant hematology and oncology indications
[5]. The emergence of targeted cancer drugs and immune checkpoint inhibitors was accompanied by increased use of AA in oncology, with 70 approvals from 2014–2018 [5]. Six oncology
AAs granted before 2009 were withdrawn, with three of the drugs resubmitted and approved
based on improved evidence supporting their use [6]. A seventh withdrawal was initiated in
2019 [7]. Almost all of the cancer AAs prior to 2016 have been converted to full approvals,
while most from 2016–2018 remain to be converted.
All stakeholders (patients, physicians, sponsors of innovative treatments, reimbursement
agencies, and regulators) ideally should benefit from AA. However, concern has been
expressed about the value of accelerated approvals [8]. Some of the main points of conflict are
over what has been measured in studies supporting accelerated approval, who has been
assessed in these studies, and how long it will take for results to become available for conversion to traditional approval. AA trials provide less information than regulatory approval usually requires, and the short-term benefit/risk balance based on these trials may not reflect the
true benefit/risk balance for the product. Decision-makers must rely on evidence that is less
complete than what current practices provide to create cost-effectiveness assessments of new
treatments [9]. Although the regulations concerning AA require sponsors to conduct clinical
trials to establish efficacy based on clinical benefit, adherence in this regard has not been universal [10]. Issues affecting the completion of these requirements include equipoise concerns
when a drug has been shown superior to standard care in the AA trials [11] and reluctance of
patients to participate in a randomized confirmatory trial of a marketed drug available without
the risks or inconvenience of research.
A common definition of the intuitively appealing term ‘impact’ that is used in discussions
of the benefits and risks of AA can be elusive. There are many possibilities, depending on the
needs and values of various stakeholders. For this discussion, impact refers to the value of suitable metrics reflecting subsequent actions that a stakeholder might take as a result of an evaluation of the information supporting AA. For example(s), a regulatory agency may decide to
approve or not approve a drug via AA depending on the distribution of possible predicted
metric values such as estimated probability of survival for more than 5 years regardless of
promising early findings. An HTA could decide to put a drug approved via AA on its formulary for reimbursement (or not) depending on the distribution of possible predicted metric
values such as whether the anticipated cost per quality-adjusted life year was within an acceptable range. A sponsor might decide whether or not to go forward with seeking AA given the
metric value findings from trials that might be submitted in support seeking AA. Physicians
and patients might decide to initiate treatment (or not) depending on possible predicted outcomes. “Values” in this context depends on what is important (and how important) to each
PLOS ONE | https://doi.org/10.1371/journal.pone.0265712 June 24, 2022
2 / 19
PLOS ONE
Impact of accelerated approval
stakeholder. Stakeholder values would be elicited in any implementation. Different stakeholders will have different attributes that are important to them, and even may differ in how importantly they regard attributes that are shared with other stakeholders.
An evaluation of the impact of AA is essential for assessing its value as it applies to the
stakeholders that it affects. Values for stakeholders can be unique or shared, as illustrated conceptually in Fig 1. A common process for evaluating the impact of AA can be helpful in integrating the assessments reflecting the values for the various stakeholders. These values can
provide useful guidance for decisions that any stakeholder might consider.
A recent white paper from the Institute for Clinical and Economic Review [12] proposes
strategies for strengthening the performance of the AA process by addressing a number of
concerns that are relevant for stakeholder assessment once AA has been granted. These concerns include inconsistencies in the level of uncertainty deemed to qualify surrogate endpoints
as reasonably likely to predict a clinically meaningful treatment effect; a lack of clarity over
Fig 1. Unique and shared values of key AA program stakeholders.
https://doi.org/10.1371/journal.pone.0265712.g001
PLOS ONE | https://doi.org/10.1371/journal.pone.0265712 June 24, 2022
3 / 19
PLOS ONE
Impact of accelerated approval
what magnitude of change in such outpoints justifies accelerated approval; and the high prices
commanded by these products despite their relative lack of evidence.
We propose a common framework and, therefore, a common language, for addressing the
challenges of decision making faced by various stakeholders inherent with the Accelerated
Approval process that differs from decision science based methods that have been described
previously. With this proposal, each stakeholder uses information from the AA trials to decide
on a course of action that reflects stakeholder-specific values by an objective, transparent, and
reproducible process that differs among stakeholders only with respect to the stakeholder-specific actions and values. This approach provides a way for stakeholders to understand the bases
for each other’s decisions and may be useful for informing their own decisions. The specificity
of the values for different stakeholders means that the stakeholders could reach different action
decisions, possibly at different points in time, so that there may be no immediate determination of ‘consensus’. Consequently, the common framework approach described here differs
from the consensus-based decision model for assessing health systems described by Xu et al
[13] in which different statistical decision methods employed by stakeholders with a common
objective are integrated to reach a consensus. It also differs from previous applications of decision science methods that focus on sponsor decisions about confirmatory or post-POC study
designs, and how these design options could affect outcomes, utility, and future (postapproval) decisions by stakeholders [14–18].
The context differs as well from previous work in that not all stakeholders implement their
decision processes contemporaneously. Accelerated Approval trials provide much drugrelated (and competitive) evidence that can affect some stakeholder decisions. This evidence
includes the study results, label content, FDA review documents (sometimes), and the current
treatment landscape. Most importantly, since the drug has been approved to market (at least
in the US), the decisions at issue are pushed from the sponsor and regulatory agencies to stakeholders such as patients, physicians, and health technology agencies that determine implementation in clinical care.
Method
Defining impact
This article outlines a process for defining ‘impact’ that is grounded in well-established decision theory. Although the process is objective and reproducible, its components may be objective or subjective, to reflect clinical reality, and the implementation details can be adapted to
the needs of different stakeholders. For clarity of exposition, the description that follows
focuses on the impact of AA on patients who receive the drug through access enabled by accelerated approval. Similar developments can be carried out for the other stakeholders. Decision
theoretic concepts reflecting the values of different stakeholders have been applied to the
design and analysis of clinical trials when only a subset of the patient population may be likely
to benefit materially from the test drug [14, 17–19].
There are at least two key assumptions and two parts to the process. The first key assumption is that the expected clinical benefit of a treatment for a patient is a function of the shortterm outcomes observed in the studies supporting AA, observable patient attributes, and, possibly, unobserved confounding factors. The assumption implies that the short-term benefits
on biomarkers or surrogate endpoints demonstrated by these studies would predict reliably
(in a probabilistic sense) the actual clinical benefits and risks for a patient undergoing longterm treatment. This could be questionable for various reasons, especially when there is a substantial lapse of time between the accelerated and the clinical endpoints. The anticipated clinical benefit generally will depend on the nature of the disease itself and the expectations of the
PLOS ONE | https://doi.org/10.1371/journal.pone.0265712 June 24, 2022
4 / 19
PLOS ONE
Impact of accelerated approval
treatment, whether short-term or long-term, including the usually disease-specific therapeutic
strategy.
The second key assumption, common to all clinical development programs, is that patients
are partially or wholly exchangeable [20] so that the same model applies at least for important
subgroups of the patients (e.g., men, women), certainly for those in the AA trials and ideally
for all patients who might be treated.
The first part of the process entails estimation of the parameters of models relating clinical
responses/outcomes resulting from the innovative treatment to the patient attributes and short
term findings from AA studies, including assessments of the variability of the parameter estimates and the uncertainty about predicted clinical outcomes. Ideally, if (a) the general functional (mathematical) relationship between the attributes of a patient and the clinical response
is known, (b) the exact (error-free) values of the parameters that characterize the functional
relationship as it applies to a particular patient are known, and (c) there are no confounding
factors then, at least in principle, it would be possible to determine the expected clinical outcome for any patient. In reality, the model and parameter values are not known exactly,
although reasonable approximations may be achievable based on available data. However,
even if one could determine the expected clinical outcome exactly for any patient, intrinsic variability among patients would lead to different actual responses of patients with the same
expected response, so that parameter values still would need to be estimated.
The second part of the process is the use of the parameter estimates to predict the efficacy
and toxicity outcome (clinical response) for a patient who is a candidate for treatment with a
product that received AA. This prediction is subject to uncertainty due to intrinsic patient variability, the uncertainty associated with the estimates of the parameters used to determine the
expected response, and the influence of possible confounders.
Evaluations of the impact of AA do not take place in an innovative vacuum. The overall
benefit of a therapy in a given population will depend on how long it remains in use before it is
replaced by superior therapeutic alternatives [19, 21]. The therapeutic alternatives at the time
of accelerated approval will influence the initial impact of the AA, while subsequent approvals
of other therapies for the same indication and patient population will impact the overall time
course of this impact.
Impact on patients
The potential impact for a patient treated with a product receiving AA could be positive if the
product conveys clinical benefit relative to the best available treatment option for the patient,
or negative if it prevents the patient from receiving a better treatment or makes the patient’s
condition worse because of toxicity. Many other factors influence the impact on a patient, particularly those affecting quality of life such as discomfort, management of dosing, lost work
hours, burdensome medical costs, and paperwork management. These can in principle be
incorporated in the assessment of impact, although specifically how would depend on the particular circumstances. Not treating the patient with a truly effective product could have a negative impact because the patient would fail to achieve the product’s benefit while avoiding
potential toxicities of alternative treatments. Ideally, AA provides information for estimating
the probabilities corresponding to possible therapeutic outcomes for a patient that can be
refined as more information accumulates. Eichler et al [22] describe how this might be carried
out.
An unintended negative impact of granting AA could be the failure of an ongoing trial of a
potentially effective agent because of patient reluctance to be randomized so that future
patients may not receive the drug because a reimbursement body (e.g. IQWiG [Institut für
PLOS ONE | https://doi.org/10.1371/journal.pone.0265712 June 24, 2022
5 / 19
PLOS ONE
Impact of accelerated approval
Qualität und Wirtschaftlichkeit im Gesundheitswesen] in Germany) could conclude that
insufficient evidence of benefit had been shown. Consequently, it could be more beneficial to
refuse an AA application if full approval could be granted shortly afterwards when longer term
data are available.
A key objective of this article is the proposal to assess the impact of AA in an objective,
transparent, quantitative, and reproducible way by means of well-known statistical decision
principles. S2 Appendix briefly outlines the statistical decision process used here; detailed
descriptions of the theory and principles are available in the literature [23–26]. Realistic models could incorporate more detailed specification of what might be meant by ‘outcome’, effects
of possible selection bias based on promising early results, the choice of endpoints, consideration of the influence of possible differences between the populations in the AA trials, in the
confirmatory trials. and in general clinical practice. At best, caution is required because the
findings from one trial may not be replicated in a second trial or in a general clinical population [27–29].
Considerations for assessing impact
The efficacy and toxicity outcomes of a therapy are important considerations for assessing the
impact of AA, but not the only ones; different stakeholders may focus on different issues such
as benefit/risk or cost considerations. What follows focuses for simplicity of exposition on efficacy and toxicity issues. Suppose for example, that the short-term efficacy information from
the AA trials defines a predictive distribution of a relevant clinical response (e.g., progressionfree survival past 6 months), and that a similar distribution of clinical response can be determined for a previous therapy. Fig 2 displays some possibilities for the distributions of the clinical response for the new and previous therapies, including the possibility that the distribution
of clinical responses for the new therapy may be reflect greater variability than for the previous
therapy because of less information about the clinical responses for the new therapy.
The orderings of the clinical response distributions can be expressed by various metrics.
Basing the orderings of efficacy (and toxicity) on the predictive distributions of the outcomes
reflects uncertainty both about the actual AA outcomes and the basis for inferring the likely
clinical outcomes based on the AA results.
Suppose that the efficacy information provided in support of AA uses patient and treatment
attributes to map the predicted efficacy outcomes (E = [E1, E2, E3]) to the categories/ conclusions illustrated in Fig 2. These categories refer to the desired efficacy outcomes, not to the
metrics used to support AA. The mapping depends on the indication, the disease, clinical judgment, and the expectations of therapy.
Fig 2. Orderings of clinical response for a new and a previous therapy.
https://doi.org/10.1371/journal.pone.0265712.g002
PLOS ONE | https://doi.org/10.1371/journal.pone.0265712 June 24, 2022
6 / 19
PLOS ONE
Impact of accelerated approval
Table 1. Predicted toxicity outcome.
T=
T1:
Less toxic than current therapy
T2:
About as toxic as current therapy
T3:
More toxic than current therapy, but manageable
T4:
Intolerable toxicity
https://doi.org/10.1371/journal.pone.0265712.t001
The predicted toxicity findings similarly can be assigned to various categories as in Table 1.
‘Better’, ‘outcome’, and ‘toxicity’ can mean different things to different people/stakeholders.
We focus here on their meaning from the patient’s point of view. Suppose that probabilities
associated with the various efficacy and toxicity categories can be related to predictive information about a patient who could be, but has not yet been, treated with the product, and the
condition to be treated. The prediction depends on the attributes of the candidate patient and
the parameters of the model that relates these attributes (including treatment) to the anticipated clinical outcome. The predicted outcome is subject to uncertainty about the true values
of the parameters, including the accuracy of the prediction of clinical outcome based on surrogate measures, and intrinsic variation among patients with identical attributes. This simple
approach can be extended in various ways including incorporating time into the evaluation of
the effect of AA.
These outcome categories can be combined into clinically meaningful categories describing
predicted therapeutic outcomes for patients that combine efficacy and toxicity. These categories ordinarily will be defined in numerical terms such as expected duration of progressionfree survival so that, for instance, a better outcome could be a lower confidence bound on the
hazard ratio for the new treatment relative to the standard that exceeded 1.5 (i.e., 50%
improvement in survival). What follows assumes that there are appropriate definitions of categories of efficacy and toxicity. Let yP denote the predictive observations made on a candidate
for treatment with the product in question. The surrogate endpoint and biomarker values
obtained in the AA trials may be useful for determining predictive distributions of therapeutic
outcomes and, consequently, estimates of the probabilities associated with each of the Efficacy
and Toxicity outcomes if the underlying, possibly disease-specific, assumptions that are
required can be satisfied. Table 2 lists the therapeutic outcome categories, the corresponding
true probabilities, and the probabilities estimated from the AA findings. The potentially 12 categories corresponding to the possible combinations of efficacy and toxicity outcomes are collapsed into 6 because materially less efficacy, intolerable toxicity, or worse toxicity without
materially better efficacy all identify unacceptable clinical outcomes that do not need to be
considered separately.
The estimated probability of the last category [w6(yP)] may be low because any approval,
let alone AA, seems unlikely in that case. However, reality may be different if the intolerable
toxicity is a rare event unobserved at the time of AA. The probabilities associated with the
Table 2. True and estimated probabilities associated with the therapeutic outcomes in Fig 2 and Table 1 for a patient with attributes yP.
True probabilities
AA estimated probabilities
Toxicity relative to current therapy
Toxicity relative to current therapy
Efficacy relative
Less
Similar
Worse
Intolerable
Less
Similar
Worse
to current therapy
T1
T2
T3
T4
T1
T2
T3
E1
W1(yP)
W2(yP)
W3(yP)
w1(yP)
w2(yP)
w3(yP)
Similar
E2
W4(yP)
W5(yP)
w4(yP)
w5(yP)
Inferior
E3
Materially better
W6(yP)
Intolerable
T4
w6(yP)
https://doi.org/10.1371/journal.pone.0265712.t002
PLOS ONE | https://doi.org/10.1371/journal.pone.0265712 June 24, 2022
7 / 19
PLOS ONE
Impact of accelerated approval
combined outcomes generally will depend on both the efficacy and toxicity outcomes. Efficacy
and toxicity may reflect a common mechanism of action, and both are often correlated with
pharmacokinetic exposure.
Statistical considerations
Details of how the statistical issues should be addressed are outside the scope of this paper
because they will be situation-specific. Nonetheless, sources of uncertainty should be kept in
mind when using short-term AA findings based on biomarkers and surrogate measures to predict clinical outcomes. Factors affecting the clinical outcome relevance and predictability for
patients include
1. Early trial sampling variance, e.g., the standard error of the early endpoints from AA trials.
2. Selection bias that occurs because AA ordinarily would be granted only when early findings
are strong. Some drugs of modest efficacy could by chance provide an early finding strong
enough to justify AA, while some drugs of greater efficacy could not provide sufficiently
strong early findings, also by chance, and thereby not be granted AA.
3. The correlation between the observations that justify AA and actual clinical outcomes–that
is, how well the early findings predict clinical outcome.
4. How well the population represented in the AA trials represents the general population for
whom the drug may be prescribed and the circumstances of its clinical use.
5. The consistency of a patient’s predicted clinical outcome from treatment that reflects the
variability of the attributes used to predict the outcome and the randomness of the patient’s
outcome given a set of attribute values.
6. The variability and sensitivity of the measures of utility to the prior specifications of uncertainty and the effect of the amount of information used for their computation.
7. Decision makers have unequal access to information. Sponsors and FDA have access to
details of individual patients in the drug development programs, while other stakeholders
(patients, prescribers, HTAs, and payers) do not have this information.
8. Small sample sizes: AA trials tend to include fewer patients than conventional regulatory
trials. Consequently, their findings are subject to more uncertainty and may present an
overly optimistic picture of treatment effect.
For example, as illustrated in Fig 3, an analysis of 24 month survival in 41 trials of treatments for recurrent glioblastoma found that the smallest trials had the highest survival rates
while the larger trials tended to have lower survival rates [30].
Determining the values of the consequences of stakeholder actions
The discussion to this point, summarized in Table 2, describes a way to conceptualize the
probabilities associated with the clinical outcomes based on the information provided in the
AA trials. However, these probabilities do not by themselves provide sufficient guidance for
determining the impact of these trials. This impact depends on how information from the trials drives decisions about possible treatment actions. These decisions also depend on the values the stakeholder attaches to the therapeutic outcome possibilities in Table 2. A numerical
measure that reflects these probabilities and values provides a quantitative way to assess
impact. This approach is outlined below, and a hypothetical example is presented in the next
section.
PLOS ONE | https://doi.org/10.1371/journal.pone.0265712 June 24, 2022
8 / 19
PLOS ONE
Impact of accelerated approval
Fig 3. Proportion of patients surviving 24 months or longer in 41 trials of treatments for recurrent glioblastoma [30].
https://doi.org/10.1371/journal.pone.0265712.g003
At its most basic level, the decision process amounts to generating an action about treatment, e.g.,
a1 ¼ “Refuse treatment”; a2 ¼ “Accept treatment”; a3 ¼ “Delay=defer treatment”;
The third possibility, “Delay/defer treatment” allows for some hedging, e.g., to obtain further
information about the anticipated effect of treatment by consultation with other specialists, or
to consider further the implications about the uncertainty of the predicted outcome due to the
limited information from the AA trials. The action taken depends on the information that efficacy and toxicity findings from the AA trials provide and the anticipated relevance for the
individual patient. For the present, suppose that the action taken depends on the estimated
probabilities for the patient of the various clinical therapeutic outcomes, which may be all that
the physician and patient know from the AA findings [31, 32].
Suppose that a stakeholder can assign a value (‘utility’ in decision analysis terms, or desirability) to the consequence (therapeutic outcome) of any action that the stakeholder might
take. The term “va
