Algorithmic identification of persons with dementia for research recruitment: ethical considerations.

Underdiagnosis, misdiagnosis, and patterns of social inequality that translate into unequal access to health systems all pose barriers to identifying and recruiting diverse and representative populations into research on Alzheimer's disease and Alzheimer's disease related dementias. In response, some have turned to algorithms to identify patients living with dementia using information that is associated with this condition but that is not as specific as a diagnosis. This paper explains six ethical issues associated with the use of such algorithms including the generation of new, sensitive, identifiable medical information for research purposes without participant consent, issues of justice and equity, risk, and ethical communication. It concludes with a discussion of strategies for addressing these issues and prompting valuable research.

Digital Health Technologies in Clinical Trials: An Ontology-Driven Analysis to Inform Digital Sustainability Policies.

BACKGROUND: Digital health technologies (DHTs) can facilitate the execution of de-centralized trials that can offer opportunities to reduce the burden on participants, collect outcome data in a real-world setting, and potentially make trial populations more diverse and inclusive. However, DHTs can also be a significant source of electronic waste (e-waste). In recognition of the potential health and environmental impact from DHT use in trials, private and public institutions have recently launched initiatives to help measure and manage this e-waste. But in order to develop sound e-waste management policies, it will be necessary to first estimate the current volume of e-waste that results from the use of DHTs in trials. MATERIALS AND METHODS: A Web Ontology Language (OWL)-compliant ontology of DHTs was created using a list of 500 DHT device names derived from a mixture of public and private sources. The U.S. clinical trials registry, ClinicalTrials.gov, was then queried to identify and classify trials using any of the devices in the ontology. The ClinicalTrials.gov records from this search were then analyzed to characterize the volume and properties of trials using DHTs, as well as estimating the total volume of individual DHT units that have been provisioned (or are planned to be provisioned) for clinical research. RESULTS: Our ontology-driven search identified 2326 unique clinical trials with a reported "actual" enrollment of 200,947 participants and a "planned" enrollment of an additional 4,094,748 participants. The most-used class of DHTs in our ontology was "wearables," (1852 trials), largely driven by the use of smart watches and other wrist-worn sensors (estimated to involve 149,391 provisioned devices). The most-used subtype of DHTs in trials was "subcutaneous" devices (367 trials), driven by the prevalent use and testing of glucose monitors (estimated to involve 17,666 provisioned devices). CONCLUSION: Thousands of trials, involving hundreds of thousands of devices, have already been completed, and many more trials (potentially involving millions more devices) are planned. Despite the great opportunities that are afforded by DHTs to the clinical trial enterprise, if the industry lacks the ability to track DHT use with sufficient resolution, the result is likely to be a great deal of e-waste. A new ontology of DHTs, combined with rigorous data science methods like those described in this paper, can be used to provide better information across the industry, and in turn, help create a more sustainable and equitable clinical trials enterprise.

Changing patient preferences toward better trial recruitment: an ethical analysis.

While randomized controlled trials are essential to health research, many of these trials fail to recruit enough participants. Approaching recruitment through the lens of behavioral science can help trialists to understand influences on the decision to participate and use them to increase recruitment. Although this approach is promising, the use of behavioral influences during recruitment is in tension with the ethical principle of respect for persons, as at least some of these influences could be used to manipulate potential participants. In this paper, we examine this tension by discussing two types of behavioral influences: one example involves physician recommendations, and the other involves framing of information to exploit cognitive biases. We argue that despite the apparent tension with ethical principles, influencing trial participants through behavior change strategies can be ethically acceptable. However, we argue that trialists have a positive obligation to analyze their recruitment strategies for behavioral influences and disclose these upfront to the research ethics committee. But we also acknowledge that since neither trialists nor ethics committees are presently well equipped to perform these analyses, additional resources and guidance are needed. We close by outlining a path toward the development of such guidance.

Opioid Use Disorder Treatments: An Evidence Map.

BACKGROUND: Evidence maps are emerging data visualization of a systematic review. There are no published evidence maps summarizing opioid use disorder (OUD) interventions. AIM: Our aim was to publish an interactive summary of all peer-reviewed interventional and observational trials assessing the treatment of OUD and common clinical outcomes. METHODS: PubMed, Embase, PsycInfo, Cochrane Central Register of Clinical Trials, and Web of Science were queried using multiple OUD-related MESH terms, without date limitations, for English-language publications. Inclusions were human subjects, treatment of OUD, OUD patient or community-level outcomes, and systematic reviews of OUD interventions. Exclusions were laboratory studies, reviews, and case reports. Two reviewers independently scanned abstracts for inclusion before coding eligible full-text articles by pre-specified filters: research design, study population, study setting, intervention, outcomes, sample size, study duration, geographical region, and funding sources. RESULTS: The OUD Evidence Map (https://med.nyu.edu/research/lee-lab/research/opioid-use-disorder-treatment-evidence-map) identified and assessed 12,933 relevant abstracts through 2020. We excluded 9455 abstracts and full text reviewed 2839 manuscripts; 888 were excluded, 1591 were included in the final evidence map. The most studied OUD interventions were methadone (n = 754 studies), buprenorphine (n = 499), and naltrexone (n = 134). The most common outcomes were heroin/opioid use (n = 708), treatment retention (n = 557), and non-opioid drug use (n = 368). Clear gaps included a wider array of opioid agonists for treatment, digital behavioral interventions, studies of OUD treatments in criminal justice settings, and overdose as a clinical outcome. CONCLUSION: This OUD Evidence Map highlights the importance of pharmacologic interventions for OUD and reductions in opioid use. Future iterations will update results annually and scan policy-level interventions.

Ontology-based categorization of clinical studies by their conditions.

OBJECTIVE: The free-text Condition data field in the ClinicalTrials.gov is not amenable to computational processes for retrieving, aggregating and visualizing clinical studies by condition categories. This paper contributes a method for automated ontology-based categorization of clinical studies by their conditions. MATERIALS AND METHODS: Our method first maps text entries in ClinicalTrials.gov's Condition field to standard condition concepts in the OMOP Common Data Model by using SNOMED CT as a reference ontology and using Usagi for concept normalization, followed by hierarchical traversal of the SNOMED ontology for concept expansion, ontology-driven condition categorization, and visualization. We compared the accuracy of this method to that of the MeSH-based method. RESULTS: We reviewed the 4,506 studies on Vivli.org categorized by our method. Condition terms of 4,501 (99.89%) studies were successfully mapped to SNOMED CT concepts, and with a minimum concept mapping score threshold, 4,428 (98.27%) studies were categorized into 31 predefined categories. When validating with manual categorization results on a random sample of 300 studies, our method achieved an estimated categorization accuracy of 95.7%, while the MeSH-based method had an accuracy of 85.0%. CONCLUSION: We showed that categorizing clinical studies using their Condition terms with referencing to SNOMED CT achieved a better accuracy and coverage than using MeSH terms. The proposed ontology-driven condition categorization was useful to create accurate clinical study categorization that enables clinical researchers to aggregate evidence from a large number of clinical studies.

An analysis of published trials found that current use of pragmatic trial labels is uninformative.

OBJECTIVES: Randomized trials labelled as "pragmatic" are attractive to funders, patients, and clinicians as the label implies that the results are directly applicable to clinical care. We examined how authors justify use of the label (e.g., by referring to one or more PRECIS [PRagmatic Explanatory Continuum Indicator Summary]-2 domains). STUDY DESIGN AND SETTING: We reviewed primary trial reports published 2014-2019, registered in ClinicalTrials.gov and using the pragmatic label anywhere in the report. RESULTS: Among 415 trials, the label was justified by reference to at least one design element in 282 (68.0%); of these, 240 (85.1%) referenced trial characteristics that can be mapped to one or more of the PRECIS-2 domains, most commonly eligibility (91, 32.3%), setting (90, 31.9%), flexibility delivery (89, 31.6%), and organization (75, 26.6%); 42 (14.9%) referenced characteristics that are not PRECIS-2 domains, most commonly type of intervention/comparator (48, 17%), recruitment without consent (22, 7.8%), routinely collected data (22, 7.8%), and cluster randomization (20, 7.1%). Most reports referenced only one or two design elements. Overall, 9/415 (2.2%) provided PRECIS wheels. CONCLUSION: Current use of pragmatic labels is uninformative. Authors should clarify the decision the trial is intended to support and include a PRECIS-2 table to make the design transparent.

Ethical considerations within pragmatic randomized controlled trials in dementia: Results from a literature survey.

Introduction: This review aims to describe the landscape of pragmatic randomized controlled trials (RCTs) in the context of Alzheimer's disease (AD) and related dementias with respect to ethical considerations. Methods: Searches of MEDLINE were performed from January 2014 until April 2019. Extracted information included: trial setting, interventions, data collection, study population, and ethical protections (including ethics approvals, capacity assessment, and informed consent). Results: We identified 62 eligible reports. More than two-thirds (69%) included caregivers or health-care professionals as research participants. Fifty-eight (94%) explicitly identified at least one vulnerable group. Two studies did not report ethics approval. Of 57 studies in which patients were participants, 55 (96%) reported that consent was obtained but in 37 studies (67%) no mention was made regarding assessment of the patients' capacity to consent to research participation. Discussion: Few studies reported protections implemented when vulnerable participants were included. Shortcomings remain when reporting consent approaches and capacity assessment.

Principal investigators over-optimistically forecast scientific and operational outcomes for clinical trials.

OBJECTIVE: To assess the accuracy of principal investigators' (PIs) predictions about three events for their own clinical trials: positivity on trial primary outcomes, successful recruitment and timely trial completion. STUDY DESIGN AND SETTING: A short, electronic survey was used to elicit subjective probabilities within seven months of trial registration. When trial results became available, prediction skill was calculated using Brier scores (BS) and compared against uninformative prediction (i.e. predicting 50% all of the time). RESULTS: 740 PIs returned surveys (16.7% response rate). Predictions on all three events tended to exceed observed event frequency. Averaged PI skill did not surpass uninformative predictions (e.g., BS = 0.25) for primary outcomes (BS = 0.25, 95% CI 0.20, 0.30) and were significantly worse for recruitment and timeline predictions (BS 0.38, 95% CI 0.33, 0.42; BS = 0.52, 95% CI 0.50, 0.55, respectively). PIs showed poor calibration for primary outcome, recruitment, and timelines (calibration index = 0.064, 0.150 and 0.406, respectively), modest discrimination in primary outcome predictions (AUC = 0.76, 95% CI 0.65, 0.85) but minimal discrimination in the other two outcomes (AUC = 0.64, 95% CI 0.57, 0.70; and 0.55, 95% CI 0.47, 0.62, respectively). CONCLUSION: PIs showed overconfidence in favorable outcomes and exhibited limited skill in predicting scientific or operational outcomes for their own trials. They nevertheless showed modest ability to discriminate between positive and non-positive trial outcomes. Low survey response rates may limit generalizability.

Informed consent in pragmatic trials: results from a survey of trials published 2014-2019.

OBJECTIVES: To describe reporting of informed consent in pragmatic trials, justifications for waivers of consent and reporting of alternative approaches to standard written consent. To identify factors associated with (1) not reporting and (2) not obtaining consent. METHODS: Survey of primary trial reports, published 2014-2019, identified using an electronic search filter for pragmatic trials implemented in MEDLINE, and registered in ClinicalTrials.gov. RESULTS: Among 1988 trials, 132 (6.6%) did not include a statement about participant consent, 1691 (85.0%) reported consent had been obtained, 139 (7.0%) reported a waiver and 26 (1.3%) reported consent for one aspect (eg, data collection) but a waiver for another (eg, intervention). Of the 165 trials reporting a waiver, 76 (46.1%) provided a justification. Few (53, 2.9%) explicitly reported use of alternative approaches to consent. In multivariable logistic regression analyses, lower journal impact factor (p=0.001) and cluster randomisation (p<0.0001) were significantly associated with not reporting on consent, while trial recency, cluster randomisation, higher-income country settings, health services research and explicit labelling as pragmatic were significantly associated with not obtaining consent (all p<0.0001). DISCUSSION: Not obtaining consent seems to be increasing and is associated with the use of cluster randomisation and pragmatic aims, but neither cluster randomisation nor pragmatism are currently accepted justifications for waivers of consent. Rather than considering either standard written informed consent or waivers of consent, researchers and research ethics committees could consider alternative consent approaches that may facilitate the conduct of pragmatic trials while preserving patient autonomy and the public's trust in research.

A review of pragmatic trials found a high degree of diversity in design and scope, deficiencies in reporting and trial registry data, and poor indexing.

OBJECTIVE: We established a large database of trials to serve as a resource for future methodological and ethical analyses. Here, we use meta-data to describe the broad landscape of pragmatic trials including research areas, identification as pragmatic, quality of trial registry data and enrolment. STUDY DESIGN AND SETTING: Trials were identified by a validated search filter and included if a primary report of a health-related randomized trial published January 2014-April 2019. Data were collated from MEDLINE, Web of Science, ClinicalTrials.gov, and full text. RESULTS: 4337 eligible trials were identified from 13,065 records, of which 1988 were registered in ClinicalTrials.gov. Research areas were diverse, with the most common being general and internal medicine; public, environmental and occupational health; and health care sciences and services. The term "pragmatic" was seldom used in titles or abstracts. Several domains in ClinicalTrials.gov had questionable data quality. We estimated that one-fifth of trials under-accrued by at least 15%. CONCLUSION: There is a need to improve reporting of pragmatic trials and quality of trial registry data. Under accrual remains a challenge in pragmatic RCTs despite calls for more streamlined recruitment approaches. The diversity of pragmatic trials should be reflected in future ethical analyses.

Reopening schools safely in the face of COVID-19: Can cluster randomized trials help?

The COVID-19 pandemic has highlighted the challenges of evidence-based health policymaking, as critical precautionary decisions, such as school closures, had to be made urgently on the basis of little evidence. As primary and secondary schools once again close in the face of surging infections, there is an opportunity to rigorously study their reopening. School-aged children appear to be less affected by COVID-19 than adults, yet schools may drive community transmission of the virus. Given the impact of school closures on both education and the economy, schools cannot remain closed indefinitely. But when and how can they be reopened safely? We argue that a cluster randomized trial is a rigorous and ethical way to resolve these uncertainties. We discuss key scientific, ethical, and resource considerations both to inform trial design of school reopenings and to prompt discussion of the merits and feasibility of conducting such a trial.

Heterogeneity of antidiabetic treatment effect on the risk of major adverse cardiovascular events in type 2 diabetes: a systematic review and meta-analysis.

BACKGROUND: We explored whether clinically relevant baseline characteristics of patients with type 2 diabetes can modify the effect of glucagon-like peptide-1 receptor agonists (GLP-1 RA) or sodium-glucose cotransporter-2 inhibitors (SGLT-2i) on the risk of major adverse cardiovascular events (MACE). METHODS: We investigated Medline and EMBASE through June 2019. We included randomized clinical trials reporting the effect of GLP-1 RA or SGLT-2i on MACE in subgroups of patients with type 2 diabetes, identified through key baseline factors: established cardiovascular disease; heart failure; chronic kidney disease; uncontrolled diabetes; duration of diabetes; hypertension; obesity; age; gender and race. Hazard ratios (HRs) and 95% confidence intervals (CIs) from trials were meta-analyzed using random-effects models. RESULTS: Ten trials enrolling 89,790 patients were included in the analyses. Subgroup meta-analyses showed a 14% risk reduction of MACE in patients with established cardiovascular disease [GLP1-RA: HR, 0.86 (95% CI, 0.80-0.93); SGLT-2i: 0.86 (0.80-0.93)], and no effect in at-risk patients without history of cardiovascular events [GLP1-RA: 0.94 (0.82-1.07); SGLT-2i: 1.00 (0.87-1.16)]. We observed a trend toward larger treatment benefits with SGLT-2i among patients with chronic kidney disease [0.82 (0.69-0.97)], and patients with uncontrolled diabetes for both GLP1-RA or SGLT-2i [GLP1-RA: 0.82 (0.71-0.95); SGLT-2i: 0.84 (0.75-0.95)]. Uncontrolled hypertension, obesity, gender, age and race did not appear to modify the effect of these drugs. CONCLUSIONS: In this exploratory analysis, history of cardiovascular disease appeared to modify the treatment effect of SGLT2i or GLP1-RA on MACE. Chronic kidney disease and uncontrolled diabetes should be further investigated as potential effect modifiers.

Ethical and Regulatory Issues for Embedded Pragmatic Trials Involving People Living with Dementia.

Embedded pragmatic clinical trials (ePCTs) present an opportunity to improve care for people living with dementia (PLWD) and their care partners, but they also generate a complex constellation of ethical and regulatory challenges. These challenges begin with participant identification. Interventions may be delivered in ways that make it difficult to identify who is a human subject and therefore who needs ethical and regulatory protections. The need for informed consent, a core human subjects protection, must be considered but can be in tension with the goals of pragmatic research design. Thus it is essential to consider whether a waiver or alteration of informed consent is justifiable. If informed consent is needed, the question arises of how it should be obtained because researchers must acknowledge the vulnerability of PLWD due in part to diminished capacity and also to increased dependence on others. Further, researchers should recognize that many sites where ePCTs are conducted will be unfamiliar with human subjects research regulations and ethics. In this report, the Regulation and Ethics Core of the National Institute on Aging Imbedded Pragmatic Alzheimer's disease (AD) and AD-related dementias (AD/ADRD) Clinical Trials (IMPACT) Collaboratory discusses key ethical and regulatory challenges for ePCTs in PLWD. A central thesis is that researchers should strive to anticipate and address these challenges early in the design of their ePCTs as a means of both ensuring compliance and advancing science. J Am Geriatr Soc 68:S37-S42, 2020.

Evaluating the evidence behind the surrogate measures included in the FDA's table of surrogate endpoints as supporting approval of cancer drugs.

BACKGROUND: In July 2018, the FDA first published a table listing all surrogate measures that it has used, and may accept for future use, in regulatory approval. However, the strength of surrogacy for those measures was not formally assessed. Using the case example of breast cancer, we aimed to evaluate the strength of correlation of surrogate measures listed in the FDA's Table with overall survival. METHODS: This cross-sectional study of the FDA's Table of Surrogate Endpoints was conducted in May 2019. All surrogate measures listed in the FDA table as appropriate for accelerated or regular approval for breast cancer were extracted. We identified studies evaluating the correlation of treatment benefit in the surrogate with treatment benefit in overall survival and extracted results from the correlation analysis. FINDINGS: Five surrogate endpoints were listed for breast cancer in the FDA website: pathological complete response rates (pCR), event-free survival (EFS), disease-free survival (DFS), objective response rates (ORR), and progression-free survival (PFS), of which pCR was listed as appropriate only for accelerated approval, while the rest were considered appropriate for accelerated or regular approval. No correlation study evaluated the correlation of treatment effects on EFS with that on OS. The results from correlation studies evaluating pCR, DFS, ORR, and PFS suggest that the treatment effects on none of these surrogate measures were strongly correlated with treatment effects on OS (r<0.85 or R2  < 0.7, except for DFS in HER2 positive early breast cancer (R2  = 0.75). INTERPRETATION: Using breast cancer as an example, we evaluated the underlying evidence for the surrogate endpoints for solid tumors listed in the FDA's Table of Surrogate Endpoints and found weak or missing correlations of treatment effects on these surrogates with treatment effects on OS . Surrogate measures should be predictive of clinical benefit to be useful in supporting regular FDA approval. FUNDING: Work on this project was funded by the Arnold Ventures. Dr. Kesselheim is also supported by the Harvard-MIT Center for Regulatory Science. The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Cluster over individual randomization: are study design choices appropriately justified? Review of a random sample of trials.

BACKGROUND: Novel rationales for randomizing clusters rather than individuals appear to be emerging from the push for more pragmatic trials, for example, to facilitate trial recruitment, reduce the costs of research, and improve external validity. Such rationales may be driven by a mistaken perception that choosing cluster randomization lessens the need for informed consent. We reviewed a random sample of published cluster randomized trials involving only individual-level health care interventions to determine (a) the prevalence of reporting a rationale for the choice of cluster randomization; (b) the types of explicit, or if absent, apparent rationales for the use of cluster randomization; (c) the prevalence of reporting patient informed consent for study interventions; and (d) the types of justifications provided for waivers of consent. We considered cluster randomized trials for evaluating exclusively the individual-level health care interventions to focus on clinical trials where individual randomization is only theoretically possible and where there is a general expectation of informed consent. METHODS: A random sample of 40 cluster randomized trials were identified by implementing a validated electronic search filter in two electronic databases (Ovid MEDLINE and Embase), with two reviewers independently extracting information from each trial. Inclusion criteria were the following: primary report of a cluster randomized trial, evaluating exclusively an individual-level health care intervention, published between 2007 and 2016, and conducted in Canada, the United States, European Union, Australia, or low- and middle-income country settings. RESULTS: Twenty-five trials (62.5%, 95% confidence interval = 47.5%-77.5%) reported an explicit rationale for the use of cluster randomization. The most commonly reported rationales were those with logistical or administrative convenience (15 trials, 60%) and those that need to avoid contamination (13 trials, 52%); five trials (20%) were cited rationales related to the push for more pragmatic trials. Twenty-one trials (52.5%, 95% confidence interval = 37%-68%) reported written informed consent for the intervention, two (5%) reported verbal consent, and eight (20%) reported waivers of consent, while in nine trials (22.5%) consent was unclear or not mentioned. Reported justifications for waivers of consent included that study interventions were already used in clinical practice, patients were not randomized individually, and the need to facilitate the pragmatic nature of the trial. Only one trial reported an explicit and appropriate justification for waiver of consent based on minimum criteria in international research ethics guidelines, namely, infeasibility and minimal risk. CONCLUSION: Rationales for adopting cluster over individual randomization and for adopting consent waivers are emerging, related to the need to facilitate pragmatic trials. Greater attention to clear reporting of study design rationales, informed consent procedures, as well as justification for waivers is needed to ensure that such trials meet appropriate ethical standards.

A search filter to identify pragmatic trials in MEDLINE was highly specific but lacked sensitivity.

OBJECTIVES: Identifying pragmatic trials from among all randomized trials is challenging because of inconsistent reporting. Our objective was to develop and validate a search filter to identify reports of pragmatic trials from Ovid MEDLINE. STUDY DESIGN AND SETTING: Two sets of known and probable pragmatic trial records were analyzed using text mining to generate candidate terms. Two large population sets comprising clinical trials and explanatory trials were used to select discriminating terms. Various combinations of terms were tested iteratively to achieve optimal search performance. Two externally derived sets were used to validate sensitivity and specificity of the derived filters. RESULTS: Our validated sensitivity-maximizing filter (combines trial design terms with terms relating to attributes of pragmatic trials) retrieves over 42,000 records in MEDLINE and has sensitivity of 46.4% (95% confidence interval (CI) 37.2 to 55.7%) and estimated specificity of 98.1% (95% CI 93.4 to 99.8%). Search performance is superior to other ad hoc filters for pragmatic trials. The Cochrane search for randomized trials has much better sensitivity (98.2%), but poorer specificity (1.9%) and retrieves 4.5 million records. CONCLUSION: A highly specific filter (low false positive rate) with moderate sensitivity is available for identifying reports of trials more likely to be pragmatic.

The importance of decision intent within descriptions of pragmatic trials.

OBJECTIVE: It is now more than 50 years since the concepts of explanatory and pragmatic attitudes toward trials were first discussed by Schwartz and Lellouch in their influential 1967 paper. Since then, there has been increasing focus on design aspects that may be consistent with more pragmatic attitudes within clinical trials, and a number of tools developed to assist investigators prospectively think about their trial design. Researchers have subsequently expressed interest in using these tools retrospectively to characterize trials as pragmatic or explanatory. RESULTS: We suggest that recent attempts to retrospectively dichotomize trials solely on the basis of quantitative scoring of trial design features are flawed. Instead, we argue that there is a need to consider both the intent and design when assessing the degree of pragmatism within a trial. CONCLUSION: The practical implication of our suggestion for trial reporting is that investigators should explicitly state the intent of the trial through a clear articulation of the decision that they hope will be informed by the trial results. This should be coupled with a completed PRagmatic-Explanatory Continuum Indicator Summary 2 assessment (or similar) with an explanation of study design choices to appropriately assess whether the study design is consistent with the study intent. We believe this will assist reviewers and knowledge users in making assessments of trials.

The evidence landscape in precision medicine.

Precision medicine is beginning to make an impact on the treatment of different diseases, but there are still challenges that must be overcome, such as the complexity of interventions, the need for marker validation, and the level of evidence necessary to demonstrate effectiveness. In this Perspective, we describe how evidence landscapes can help to address these challenges.

A Systematic Review and Meta-Analysis of Bevacizumab in First-Line Metastatic Breast Cancer: Lessons for Research and Regulatory Enterprises.

BACKGROUND: The US Food and Drug Administration's accelerated approval and later withdrawal of bevacizumab in patients with metastatic breast cancer (mBC) is a seminal case for ongoing debates about the validity of using progression-free survival (PFS) as a surrogate measure for overall survival (OS) in cancer drug approvals. We systematically reviewed and meta-analyzed the evidence around bevacizumab's regulatory approval and withdrawal in mBC. METHODS: We searched for all published phase II or III clinical trials testing bevacizumab as a first-line therapy for patients with mBC. Data were extracted on trial demographics, interventions, and outcomes. Descriptive analysis was stratified by whether the trial was initiated before, during, or after the accelerated approval. We used a cumulative random-effects meta-analysis to assess the evolution of evidence of the effect of bevacizumab on PFS and OS. We estimated the association between the trial-level PFS and OS effect using a nonlinear mixed-regression model. RESULTS: Fifty-two studies were included. Trial activity dramatically dropped after the accelerated approval was withdrawn. Eight clinical trials reported hazard ratios (hazard ratios) and were meta-analyzed. The cumulative hazard ratio for PFS was 0.72 (95% CI = 0.65 to 0.79), and the cumulative hazard ratio for OS was 0.90 (95% CI = 0.80 to 1.01). The regression model showed a statistically nonsignificant association between PFS benefit and OS benefit (ß = 0.43, SE = 0.81). CONCLUSION: The US Food and Drug Administration's decision-making in this case was consistent with the evolving state of evidence. However, the fact that seven clinical trials are insufficient to conclude validity (or lack thereof) for a trial-level surrogate suggests that it would be more efficient to conduct trials using the more clinically meaningful endpoints.