Download Free Medical Informatics support for precision medicine PowerPoint Presentation

Informatics support for precision medicine PowerPoint Presentation

DOWNLOAD MEDICAL POWERPOINT PRESENTATION

About This Presentation

Description : Check out this medical presentation on Precision Medicine, which is titled "Informatics support for precision medicine", to know about precision medicine, informatics challenges to precision medicine, and what does it take to “practice” precision medicine.

Tags : Precision Medicine, Informatics support, informatics challenges

Published on : Jul 20, 2021 | Views : 2050 | Downloads : 0

Shareable Url :

Embed Code :

Comment

Transcript

Slide 1-

Informatics support for precision medicine Elmer V. Bernstam, MD Reynolds and Reynolds Professor of Clinical Informatics Associate Dean for Research, School of Biomedical Informatics Professor of Internal Medicine, Medical School The University of Texas Health Science Center at Houston Elmer.V.Bernstam@uth.tmc.edu
Slide 2-

Knowing is not enough; we must apply. Willing is not enough; we must do. Johann Wolfgang von Goethe 2 This is going to be a story about decision support…
Slide 3-

In this talk Clinical perspective: must make decisions but too much data that changes too rapidly for unaided human cognition Punch line: If we want to realize the vision of personalized medicine Can’t do it by demanding more from clinicians Need “true” decision support  Informatics challenge(s) Not memory aids 3
Slide 4-

Goals today What is precision medicine? Focusing on informatics, not molecular biology What are informatics challenges to precision medicine? What does it take to “practice” precision medicine? “One off” vs. “standard of care” 4
Slide 5-

Important issues that we will not cover… Evidence supporting the clinical use of molecular information Currently mixed  ?improving Biological complexity (Systems biology) Single molecules  linear pathways  networks Non-molecular precision medicine Patient reported outcomes, “exposome,” … Not a systematic review… 5
Slide 6-

What’s new? – quantitative difference Amount of “evidence” exploding How much you have to know Don’t learn in the encounter Is not “common knowledge” Personalized medicine = dividing the population into smaller groups Non-genomic medicine: divide population into few categories (2-4) with many members based on clinical characteristics e.g., age groups, gender, exposures, histology Then individualize based on what you learn in the encounter (e.g., patient preferences) Genomic medicine: divide population into many categories with very few (1) members Based on seemingly random labels PIK3CA, EGFR, KRAS, UGT1A1, HER-2/neu, c-KIT, BRAF… 6
Slide 7-

Some terminology… Precision Personalized (Genomic or) Genomically-informed Germline vs. Somatic alteration Pharmacogenomics 7
Slide 8-

Many categories, few members Nature (2000) 8
Slide 9-

Subdividing diseases Diffuse Large B Cell Lymphoma  multiple diseases Survival correlates with gene expression
Slide 10-

From clinical perspective “Any medication will have different effects on different bodies, organs of a person, at two different time points in one person's body, and in one organ at two different time points.” Avicenna (980-1037 AD) https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5379719/ Precision medicine – not new We always individualized therapy using Patient preferences Non-genomic risk profile Smoking, exposure to asbestos Genomic risk profile Family history - may still be more informative than molecular profiling Available facilities and resources Co-morbidities: many patients have more than one problem … Burke W, Psaty BM. Personalized Medicine in the era of genomics. JAMA. 2007;298(14): 1682-4. 10
Slide 11-

11 Information overload (c. 1550)
Slide 12-

Cognitive overload 12 William W. Stead. “Recalibrating Informatics True North”. AMIA NOW, 05/27/10; William W. Stead. Beyond expert-based practice. IOM (Institute of Medicine). Evidence-based medicine and the changing nature of health care: 2007 IOM annual meeting summary, (Introduction and Overview, p. 19). Washington, DC: The National Academies Press 2008.
Slide 13-

Recurrent exponential decay curve (Long tail) 13 Chang MT, et al. Identifying recurrent mutations in cancer reveals widespread lineage diversity and mutational specificity. Nat Biotechnol. 2016 Feb;34(2):155-63. PMID: 26619011 PMID: 26619011
Slide 14-

“Humanly impossible task” – Alan Rector (1989) Mulrow C, Cook D, et al. (1997) Systematic reviews: critical links in the great chain of evidence. Annals of Internal Medicine, 126: 389-91. 14 Helping with a humanly impossible task: Integrating knowledge based systems into clinical care. Published in J Hannu and L Seppo (eds), Proceedings of the Second Scandinavian Conference on Artificial Intelligence SCAI-89. Amsterdam: IOS Press, pp 453-466
Slide 15-

15 Hospital rounds (c. 2018)
Slide 16-

Data vs. information overload Data: observations about the world E.g., 0001 0011 1010 0010 0001 0010 0000 1000 Information: meaningful data E.g., PMID: 13621204 == [record in NLM’s MEDLINE] Knowledge: justified true belief E.g., Smoking causes lung cancer Cornfield J, Haenszel W, Hammond EC, Lilienfeld AM, Shimkin MB, Wynder EL. Smoking and lung cancer: recent evidence and a discussion of some questions. J Natl Cancer Inst. 1959. Jan;22(1):173-203. 16 Bernstam EV, Smith JW and Johnson TR. What is biomedical informatics? J Biomed Inform. Feb 2010, 43:1, pp. 104-10.
Slide 17-

Data vs. information overload Computers very good at managing data Resources can assign meaning to data (information) NLM: PubMed, Genbank, … Knowledge is much harder Context, exceptions, … Does your smoking increase my risk of lung cancer? It does if we live together… Human cognitive capacity is the bottleneck 17
Slide 18-

Preventative health care: colorectal cancer screening
Slide 19-

Preventative health care: breast cancer screening
Slide 20-

Preventative health care: cervical cancer screening
Slide 21-

Preventative health care: flu vaccination https://www.cdc.gov/flu/fluvaxview/nifs-estimates-nov2017.htm, accessed March 27, 2018
Slide 22-

Why? 7.4 hours/day required to satisfy USPSTF recs Yarnall KS, Pollak KI, Østbye T, Krause KM, Michener JL. Primary care: is there enough time for prevention? Am J Public Health. 2003;93(4):635-641. 2 hours of EHR time for every hour of patient care (much of it after hours) Sinsky C, Colligan L, Li L, et al. Allocation of physician time in ambulatory practice: a time and motion study in 4 specialties. Ann Intern Med. 2016;165(11):753-760.
Slide 23-

Clinicians are not compensating We “get it right” 54.9% of the time McGlynn EA, Asch SM, Adams J, et al. The quality of health care delivered to adults in the United States. NEJM 2003; 348(26): 2635-45. Estimated average 17 years for 14% of new scientific discoveries to enter routine clinical practice Westfall J, Mold J, Fagnan L. Practice-based research – “Blue Highways” on the NIH roadmap. JAMA 2007;297:403 – 6 (citing: Balas EA, Boren SA. Yearbook of Medical Informatics: Managing Clinical Knowledge for Health Care Improvement. Stuttgart, Germany: Schattauer Verlagsgesellschaft mbH; 2000) 11% of patients with actionable alterations enroll on genotype-matched trials Challenges: Awareness of test results, available trials, awareness of trials Meric-Bernstam F, Brusco L, Shaw K, Horombe C, Kopetz S, Davies MA, Routbort M, Piha-Paul SA, Janku F, Ueno N, Hong D, De Groot J, Ravi V, Li Y, Luthra R, Patel K, Broaddus R, Mendelsohn J, Mills GB. Feasibility of Large-Scale Genomic Testing to Facilitate Enrollment Onto Genomically Matched Clinical Trials. Journal of Clinical Oncology 2015 33:25, 2753-2762 . Improvement with more data not likely. Are we satisfied with this translational efficiency? 23
Slide 24-

Summary so far… From informatics perspective, precision medicine not new Quantitative not qualitative change Unaided human cognition is not enough What is needed to practice precision medicine? 24
Slide 25-

Precision medicine infrastructure: “one off” vs “standard of care” “Project” for each patient not scalable Need infrastructure to scale precision medicine Knowledge about the world (integrated into structured) Knowledge base(s) Gene-drug association Clinical trials Decision support EHR To house patient information and deliver advice 25
Slide 26-

Precision medicine infrastructure: Knowledge about what works Data  information  knowledge Profiling (“-omics”) Front door – everyone that comes in Consent issues On demand Who decides? Biases, … Expensive In oncology, useful to profile tumor + normal tissue CLIA (expensive, required for clinical decisions) vs research 26
Slide 27-

Precision Medicine Initiative (PMI) announced by Barack Obama 2015 State of the Union address $215M total $130M to NIH (national cohort) $70M to NCI (genomic drivers of cancer) $10M to FDA (regulatory structure) $5M to ONC (interoperability) formerly the PMI Cohort Program Recruit 1M+ voluntary participants starting in 2017 Biological samples + genetic data, diet/lifestyle + EHR data 27 https://www.nih.gov/research-training/allofus-research-program, accessed March 18, 2017 Collins, FS, Varmus H. A New Initiative on Precision Medicine. N Engl J Med 2015; 372:793-79
Slide 28-

All of Us Research Program core values Participation is open to all. Participants reflect the rich diversity of the U.S. Participants are partners. Participants have access to their information. Data will be accessed broadly for research purposes. Security and privacy will be of highest importance. The program will be a catalyst for positive change in research. 28 https://www.nih.gov/AllofUs-research-program/pmi-cohort-program-announces-new-name-all-us-research-program, accessed March 18, 2017
Slide 29-

Knowledge generation pipeline AACR Project GENIE at a glance The AACR Project GENIE Consortium Cancer Discov 2017;7:818-831 ©2017 by American Association for Cancer Research
Slide 30-

Precision medicine infrastructure: EHR Widely adopted  computable patient data Not perfect, but better for decision support than paper Often contain data needed for precision medicine Gaps Standards Interoperability Completeness 30
Slide 31-

Increasing EHR capabilities 31 https://www.healthit.gov/sites/default/files/briefs/2015_hospital_adoption_db_v17.pdf accessed 6/25/2018 Decision support functionality
Slide 32-

Precision medicine infrastructure: drug-gene association Input: gene(s) Output: drugs that target gene(s) Sources Literature (MEDLINE/PubMed) [NLM] – partly structured text ClinicalTrials.gov [NLM] -- partly structured text Pharma (websites) – text, if we’re lucky Drug catalogs – structured (maybe) Humans – does not scale Books, manuscripts, primary data sets, … 32
Slide 33-

33 + ClinVar https://www.ncbi.nlm.nih.gov/clinvar/
Slide 34-

Drug – gene association 34 https://pct.mdanderson.org/#/home, accessed 2/29/2016 26 genes 791 variants (as of 3/3/2016) 95,469 page views during 41,562 visits by 22,728 unique users in its first year (through 4/15/2015)
Slide 35-

PersonalizedCancerTherapy.org 35 https://pct.mdanderson.org/#/home/PIK3CA?section=Overview, accessed 2/29/2016
Slide 36-

36 https://cpicpgx.org, accessed March 20, 2017
Slide 37-

37 https://cpicpgx.org/genes-drugs/, accessed March 20, 2017
Slide 38-

How do you create and maintain these databases? Levels of evidence No (or multiple) standards Conflicting evidence Conflicting opinion regarding the evidence What should be “clinically actionable?” …. 38
Slide 39-

How do you create and maintain these databases? Can be constructed manually Can we adapt previous informatics work? Most informatics tools designed for “static” domains E.g., ASA treats MI Rapidly changing domain Todays VUS  tomorrow’s standard of care Goal: semi-automated process Fathiamini S, et al. Automated Identification of Molecular Effects of Drugs (AIMED). J Am Med Inf Assoc, 2016 Jul;23(4):758-65. 39
Slide 40-

AIMED pipeline Fathiamini S, et al. Automated Identification of Molecular Effects of Drugs (AIMED). J Am Med Inf Assoc, 2016 Jul;23(4):758-65.
Slide 41-

So, how well does it work? 41 F2: recall > precision Fathiamini S, et al. Automated Identification of Molecular Effects of Drugs (AIMED). J Am Med Inf Assoc, 2016 Jul;23(4):758-65.
Slide 42-

But… 26% of false positives actually true positives 61% of false positives relevant Only 13% not relevant Conclusions Hard informatics problem but can get useful results Manual construction/maintenance impractical Most targeted therapies in cancer available via early-phase clinical trials 42
Slide 43-

Decision support Long history dating back to 1950s 43
Slide 44-

Decision support Inconsistent benefit Vast majority of alerts overridden Stubborn informatics problems Encoding clinical knowledge is hard Hard to define necessary and sufficient conditions Humans can deal with this, computers can not Low tolerance for error Who is responsible for the error? Statistical approaches can be brittle See: Leeds abdominal pain system http://eknygos.lsmuni.lt/springer/56/698-736.pdf (20.2.1, p.702) 44
Slide 45-

Decision support: Is this time different? Molecular data more computable than clinical data EGFR mutation vs. “severe pneumonia” More precise “guidelines” Indications for targeted therapies Most targeted therapies are in clinical trials (cancer) (Relatively) precise eligibility criteria Clinical trial alerting at some institutions Approximate – some criteria met (e.g., UTHealth) Manual – labor intensive (E.g., MD Anderson) Better databases of clinical trials ClinicalTrials.gov is a start, but not enough detail for decision support 45
Slide 46-

Precision medicine vision 46 Guideline.gov
Slide 47-

How to integrate decision support into clinical care? Institution supported Decision support = cost of doing business Institutions compete based on performance Assuming better decision support  better outcomes Pathology (or radiology) model Order test  interpretation is part of test fee funded by insurance Some other model Licensing, government, professional societies (e.g., ASCO, ACP, …) 47
Slide 48-

Recap What is precision medicine: dividing the patient population into smaller and smaller sub-populations “One off”  Routine precision medicine Problem: Cognitive overload – due in part to information overload Only viable solution: Informatics support Take home points To realize the promise of precision medicine Can’t push more information onto the clinician Need patient-specific decision support Not memory aids  Informatics challenge(s) Minimum informatics infrastructure exists with some limitations Need computable: Knowledge about clinically relevant individual variation(s) Individual patient data (EHR) Decision support: at least clinical trial alerting Supporting policy (funding) 48
Slide 49-

Acknowledgements CTSA BMI and BIG teams at the UT Health Science Center at Houston Precision Oncology Decision Support (PODS) Core Team UT MD Anderson 49 Sheikh Bin Zayed Al Nahyan Foundation, Cancer Prevention Research Institute of Texas (CPRIT) Precision Oncology Decision Support Core RP150535, Data Science and Informatics Core for Cancer Research (RP170668), NCATS grant TR000371 (Center for Clinical and Translational Sciences), TR001104 (UT Southwestern Center for Translational Medicine), the Bosarge Foundation, the MD Anderson Cancer Center Support grant (NCI P30 CA016672), the Gillson Longenbaugh Foundation and the Brown Foundation.
Slide 50-

Questions? Elmer V. Bernstam, MD Reynolds and Reynolds Professor of Biomedical Informatics Professor of Internal Medicine Associate Dean for Research, School of Biomedical Informatics Elmer.V.Bernstam@uth.tmc.edu
Slide 51-

Precision medicine in the literature(# articles/year) 51 February, 2009 April 14, 2003 January 20, 2015 25% of drug approvals (35% cancer) were targeted therapies
Slide 52-

Information overload (c. 1851) 52
Slide 53-

53 Information overload (c. 2016)
Slide 54-

What is an Actionable Gene? A genomic alteration can be considered “actionable” if it: Predicts therapy response (sensitivity or resistance) Affects the function of a cancer-related gene, and can be targeted directly or indirectly with approved or investigational therapies. Is a specific eligibility criteria for enrollment onto genotype-selected trials Has demonstrated the ability to establish diagnosis or influence prognosis Is a germline alteration that predicts drug metabolism and/or adverse effects Is a germline alteration that predicts future risk of cancer or other diseases (usually considered more “actionable” if prevention or screening with early treatment is feasible) Meric-Bernstam, JNCI 2015 Slide courtesy of Funda Meric-Bernstam, MD
Slide 55-

Level of Evidence 55 Source: Li MM, Datto M, Duncavage EJ, Kulkarni S, Lindeman NI, Roy S, et al. Standards and guidelines for the interpretation and reporting of sequence variants in cancer: a joint consensus recommendation of the Association for Molecular Pathology, American Society of Clinical Oncology, and College of American Pathologists. J Mol Diagn 2017;19(1):4e23. http: //dx.doi.org/10.1016/j.jmoldx.2016.10.002.
Slide 56-

56 https://pct.mdanderson.org/pctService/resources/imageManager/image/loe, accessed March 20, 2017
Slide 57-

What does this mean for practicing internists right now? Too much information (that changes too quickly) to practice precision general medicine Mostly investigational  some work, many don’t SOME academic centers developing groups for specific areas Pharmacogenomics (e.g., Vanderbilt) Oncology (e.g., MD Anderson) 57
Slide 58-

What does this mean for practicing internists in the near future? EHRs adapting to support clinical practice (PN 041) What Physicians Really Need from EHRs to Be Successful in a Value-Based World (4/21/2018 11:15 AM - 12:45 PM) SOME academic centers developing groups for specific areas Pharmacogenomics (e.g., Vanderbilt) Oncology (e.g., MD Anderson) LIMITED support outside the institution 58