Translating ethical and quality principles for the effective, safe and fair development, deployment and use of artificial intelligence technologies in healthcare.

OBJECTIVE: The complexity and rapid pace of development of algorithmic technologies pose challenges for their regulation and oversight in healthcare settings. We sought to improve our institution's approach to evaluation and governance of algorithmic technologies used in clinical care and operations by creating an Implementation Guide that standardizes evaluation criteria so that local oversight is performed in an objective fashion. MATERIALS AND METHODS: Building on a framework that applies key ethical and quality principles (clinical value and safety, fairness and equity, usability and adoption, transparency and accountability, and regulatory compliance), we created concrete guidelines for evaluating algorithmic technologies at our institution. RESULTS: An Implementation Guide articulates evaluation criteria used during review of algorithmic technologies and details what evidence supports the implementation of ethical and quality principles for trustworthy health AI. Application of the processes described in the Implementation Guide can lead to algorithms that are safer as well as more effective, fair, and equitable upon implementation, as illustrated through 4 examples of technologies at different phases of the algorithmic lifecycle that underwent evaluation at our academic medical center. DISCUSSION: By providing clear descriptions/definitions of evaluation criteria and embedding them within standardized processes, we streamlined oversight processes and educated communities using and developing algorithmic technologies within our institution. CONCLUSIONS: We developed a scalable, adaptable framework for translating principles into evaluation criteria and specific requirements that support trustworthy implementation of algorithmic technologies in patient care and healthcare operations.

National landscape assessment of academic medical center support for expanded access to investigational products.

Expanded access (EA) provides a pathway for the clinical use of investigational products (drugs, biologics, and medical devices) for patients who are without satisfactory therapeutic options and for whom a clinical trial is not available. Academic medical centers (AMCs) are likely to encounter EA requests, but it is unknown what support is available at these institutions for physicians seeking EA for patients. METHODS: A landscape assessment was conducted at AMCs, focused on those within the Clinical and Translational Science Awards (CTSA) consortium. RESULTS: Forty-seven responses were evaluated including 42 CTSA hubs. The large majority (43 of 47 respondents) reported using single-patient EA, while 37 reported multi-patient industry sponsored EA and 37 reported multi-patient investigator-initiated EA. Only half reported central tracking of EA requests. Support was available at 89% of sites for single-patient EA but less often for multi-patient EA. Types of support varied and were focused largely on the initial submission to the FDA. CONCLUSION: Use of and support for EA is widespread at AMCs, with support focused on single-patient requests. Gaps in support are common for activities after initial submission, such as FDA reporting and data collection.

A framework for the oversight and local deployment of safe and high-quality prediction models.

Artificial intelligence/machine learning models are being rapidly developed and used in clinical practice. However, many models are deployed without a clear understanding of clinical or operational impact and frequently lack monitoring plans that can detect potential safety signals. There is a lack of consensus in establishing governance to deploy, pilot, and monitor algorithms within operational healthcare delivery workflows. Here, we describe a governance framework that combines current regulatory best practices and lifecycle management of predictive models being used for clinical care. Since January 2021, we have successfully added models to our governance portfolio and are currently managing 52 models.

ReGARDD (Regulatory Guidance for Academic Research of Drugs and Devices): The evolution of a collaborative regional CTSA-funded forum and website for regulatory support.

Availability of trained professionals to assist researchers navigating regulatory pathways for new drug and device development is limited within academic institutions. We created ReGARDD (Regulatory Guidance for Academic Research of Drugs and Devices), a regional forum initially involving regulatory professionals from four Clinical and Translational Science Award (CTSA)-funded institutions, to build and capitalize on local expertise and to develop a regulatory guidance website geared toward academic researchers. Since 2015, members organized 15 forums covering topics such as FDA premarket submissions, gene therapy, and intellectual property for devices and therapeutics. Through user feedback, targeted surveys, and ongoing iterative processes, we refined and maintained a shared regulatory website, which reached 6000+ users in 2019. Website updates improved navigation to drug versus device topic areas, provided new educational content and videos to address commonly asked questions, and created a portal for posting upcoming training opportunities. Survey respondents rated the website favorably and endorsed expanding ReGARDD as a centralized resource. ReGARDD strengthened the regional regulatory workforce, increased regulatory efficiency, and promulgated best organizational and operational practices. Broad-scale deployment of the ReGARDD model across the CTSA consortium may facilitate the creation of a network of regional forums and reduce gaps in access to regulatory support.

ClinicalTrials.gov reporting: strategies for success at an academic health center.

The Food and Drug Administration Amendments Act of 2007 (FDAAA 2007, US Public Law 110-98) mandated registration and reporting of results for applicable clinical trials. Meeting these registration and results reporting requirements has proven to be a challenge for the academic research community. Duke Medicine has made compliance with registration and results reporting a high priority. In order to create uniformity across a large institution, a written policy was created describing requirements for clinical trials disclosure. Furthermore, a centralized resource group was formed with three full time staff members. The group not only ensures compliance with FDAAA 2007, it also acts as a resource for study teams providing hands-on support, reporting, training, and ongoing education. Intensive resourcing for results reporting has been crucial for success. Due to implementation of the institutional policy and creation of centralized resources, compliance with FDAAA 2007 has increased dramatically at Duke Medicine for both registration and results reporting. A consistent centralized approach has enabled success in the face of changing agency rules and new legislation.

Warning letters to sponsor-investigators at academic health centres - the regulatory "canaries in a coal mine".

PURPOSE: This study highlights Warning Letter (WL) findings issued to sponsor-investigators (S-Is) by the Food and Drug Administration (FDA). METHODS: The online index of WLs issued from October 1, 2007 through September 30, 2012 was reviewed [1]. Through a manual screening process, letters were evaluated if specifically issued to 'clinical investigators', 'sponsors' or 'sponsor-investigators'. A particular focus was given to S-Is at Academic Health Centres (AHCs). Each letter was scored for the presence of violations in 40 general regulatory categories. RESULTS: A review of FDA WLs issued over a five-year period (FDA Fiscal Years 2008-2012) revealed that WLs to S-Is represent half of the WLs issued to all sponsors (16 of 32 letters). A review of these letters indicates that S-Is are not aware of, or simply do not meet, their regulatory responsibilities as either investigators or sponsors. In comparing total sponsor letters to those of S-Is, the most cited violation was the same: a lack of monitoring. A review of publicly available inspection data indicates that these 16 letters merely represent the tip of the iceberg. CONCLUSION: This review of the WL database reveals the potential for serious regulatory violations among S-Is at AHCs. Recent translational funding initiatives may serve to increase the number of S-Is, especially among Academic Health Centres (AHCs) [2]; thus, AHCs must become aware of this S-I role and work to support investigators who assume both roles in the course of their research.

Cellular mechanisms controlling caspase activation and function.

Caspases are the primary drivers of apoptotic cell death, cleaving cellular proteins that are critical for dismantling the dying cell. Initially translated as inactive zymogenic precursors, caspases are activated in response to a variety of cell death stimuli. In addition to factors required for their direct activation (e.g., dimerizing adaptor proteins in the case of initiator caspases that lie at the apex of apoptotic signaling cascades), caspases are regulated by a variety of cellular factors in a myriad of physiological and pathological settings. For example, caspases may be modified posttranslationally (e.g., by phosphorylation or ubiquitylation) or through interaction of modulatory factors with either the zymogenic or active form of a caspase, altering its activation and/or activity. These regulatory events may inhibit or enhance enzymatic activity or may affect activity toward particular cellular substrates. Finally, there is emerging literature to suggest that caspases can participate in a variety of cellular processes unrelated to apoptotic cell death. In these settings, it is particularly important that caspases are maintained under stringent control to avoid inadvertent cell death. It is likely that continued examination of these processes will reveal new mechanisms of caspase regulation with implications well beyond control of apoptotic cell death.

Rsk-mediated phosphorylation and 14-3-3ɛ binding of Apaf-1 suppresses cytochrome c-induced apoptosis.

Many pro-apoptotic signals trigger mitochondrial cytochrome c release, leading to caspase activation and ultimate cellular breakdown. Cell survival pathways, including the mitogen-activated protein kinase (MAPK) cascade, promote cell viability by impeding mitochondrial cytochrome c release and by inhibiting subsequent caspase activation. Here, we describe a mechanism for the inhibition of cytochrome c-induced caspase activation by MAPK signalling, identifying a novel mode of apoptotic regulation exerted through Apaf-1 phosphorylation by the 90-kDa ribosomal S6 kinase (Rsk). Recruitment of 14-3-3ɛ to phosphorylated Ser268 impedes the ability of cytochrome c to nucleate apoptosome formation and activate downstream caspases. High endogenous levels of Rsk in PC3 prostate cancer cells or Rsk activation in other cell types promoted 14-3-3ɛ binding to Apaf-1 and rendered the cells insensitive to cytochrome c, suggesting a potential role for Rsk signalling in apoptotic resistance of prostate cancers and other cancers with elevated Rsk activity. Collectively, these results identify a novel locus of apoptosomal regulation wherein MAPK signalling promotes Rsk-catalysed Apaf-1 phosphorylation and consequent binding of 14-3-3ɛ, resulting in decreased cellular responsiveness to cytochrome c.

Restraint of apoptosis during mitosis through interdomain phosphorylation of caspase-2.

The apoptotic initiator caspase-2 has been implicated in oocyte death, in DNA damage- and heat shock-induced death, and in mitotic catastrophe. We show here that the mitosis-promoting kinase, cdk1-cyclin B1, suppresses apoptosis upstream of mitochondrial cytochrome c release by phosphorylating caspase-2 within an evolutionarily conserved sequence at Ser 340. Phosphorylation of this residue, situated in the caspase-2 interdomain, prevents caspase-2 activation. S340 was susceptible to phosphatase 1 dephosphorylation, and an interaction between phosphatase 1 and caspase-2 detected during interphase was lost in mitosis. Expression of S340A non-phosphorylatable caspase-2 abrogated mitotic suppression of caspase-2 and apoptosis in various settings, including oocytes induced to undergo cdk1-dependent maturation. Moreover, U2OS cells treated with nocodazole were found to undergo mitotic catastrophe more readily when endogenous caspase-2 was replaced with the S340A mutant to lift mitotic inhibition. These data demonstrate that for apoptotic stimuli transduced by caspase-2, cell death is prevented during mitosis through the inhibitory phosphorylation of caspase-2 and suggest that under conditions of mitotic arrest, cdk1-cyclin B1 activity must be overcome for apoptosis to occur.

Differential Apaf-1 levels allow cytochrome c to induce apoptosis in brain tumors but not in normal neural tissues.

Brain tumors are typically resistant to conventional chemotherapeutics, most of which initiate apoptosis upstream of mitochondrial cytochrome c release. In this study, we demonstrate that directly activating apoptosis downstream of the mitochondria, with cytosolic cytochrome c, kills brain tumor cells but not normal brain tissue. Specifically, cytosolic cytochrome c is sufficient to induce apoptosis in glioblastoma and medulloblastoma cell lines. In contrast, primary neurons from the cerebellum and cortex are remarkably resistant to cytosolic cytochrome c. Importantly, tumor tissue from mouse models of both high-grade astrocytoma and medulloblastoma display hypersensitivity to cytochrome c when compared with surrounding brain tissue. This differential sensitivity to cytochrome c is attributed to high Apaf-1 levels in the tumor tissue compared with low Apaf-1 levels in the adjacent brain tissue. These differences in Apaf-1 abundance correlate with differences in the levels of E2F1, a previously identified activator of Apaf-1 transcription. ChIP assays reveal that E2F1 binds the Apaf-1 promoter specifically in tumor tissue, suggesting that E2F1 contributes to the expression of Apaf-1 in brain tumors. Together, these results demonstrate an unexpected sensitivity of brain tumors to postmitochondrial induction of apoptosis. Moreover, they raise the possibility that this phenomenon could be exploited therapeutically to selectively kill brain cancer cells while sparing the surrounding brain parenchyma.

Human beta-defensin-1, a potential chromosome 8p tumor suppressor: control of transcription and induction of apoptosis in renal cell carcinoma.

Human beta-defensin-1 (hBD-1) is a candidate tumor suppressor gene located on chromosome 8p23. Previously, we showed that cancer-specific loss of hBD-1 was found in 90% of renal clear cell carcinomas and in 82% of prostate cancers. To investigate the possible mechanisms of decreased gene expression and determine the function of hBD-1 protein in urological cancers, we sequenced hBD-1 gene coding regions in prostatic and renal cancer samples. We then analyzed the frequency distribution of promoter polymorphisms and determined the effect of these base changes on transcriptional activity of the hBD-1 promoter. A polymorphism at -688 bases upstream of the ATG start codon affects hBD-1 promoter activity, leading to a rate of reporter gene transcription that is 40% to 50% lower than the wild-type sequence when tested in either DU145 or TSU-Pr1 cell lines. In addition, a polymorphism at -44 bases was shown to enhance transcription up to 2.3 times more than the wild-type sequence in the same cell lines. In addition, three novel hBD-1 promoter mutations were found in renal and prostate cancer clinical samples. An iso-5-aza-2'-deoxycytidine treatment was effective in transcription up-regulation in DU145, suggesting a possible upstream methylation-dependent effect. Synthetic hBD-1 peptide inhibited bladder cancer cell TSU-Pr1 proliferation. Overexpression of the hBD-1 gene in renal cancer cells SW156 resulted in caspase-3-mediated apoptosis. These data support the hypothesis that hBD-1 is a potential tumor suppressor gene for urological cancers. Promoter point mutations may be responsible for cancer-specific loss of hDB-1 expression.

Enhanced sensitivity to cytochrome c-induced apoptosis mediated by PHAPI in breast cancer cells.

Apoptotic signaling defects both promote tumorigenesis and confound chemotherapy. Typically, chemotherapeutics stimulate cytochrome c release to the cytoplasm, thereby activating the apoptosome. Although cancer cells can be refractory to cytochrome c release, many malignant cells also exhibit defects in cytochrome c-induced apoptosome activation, further promoting chemotherapeutic resistance. We have found that breast cancer cells display an unusual sensitivity to cytochrome c-induced apoptosis when compared with their normal counterparts. This sensitivity, not observed in other cancers, resulted from enhanced recruitment of caspase-9 to the Apaf-1 caspase recruitment domain. Augmented caspase activation was mediated by PHAPI, which is overexpressed in breast cancers. Furthermore, cytochrome c microinjection into mammary epithelial cells preferentially killed malignant cells, suggesting that this phenomenon might be exploited for chemotherapeutic purposes.