Why was the cohort set up?
Genetic modulation of drug response can cause life-threatening adverse drug reactions, increase susceptibility to drug–drug interactions and alter therapeutic effectiveness.1–3 Pharmacogenomics has the potential to immediately impact the care of patients and is a hallmark of genomic medicine worldwide. Ideally, pharmacogenomics-guided drug selection will produce optimal effects for specific indications, reduce delays in care to effective therapies, improve patient safety and reduce health care costs. Over the past decade, a large number of pharmacogenomic variants with demonstrated clinical utility have been identified and corresponding international guidelines for clinical implementation have been developed and published by the Clinical Pharmacogenetics Implementation Consortium (CPIC),4 Canadian Pharmacogenomics Network for Drug Safety (CPNDS)5 and Royal Dutch Association for the Advancement of Pharmacy Dutch Pharmacogenetics Working Group (DPWG).6 There is now general agreement that pharmacogenomics represents an area within genomic science that could have a significant positive impact on clinical medicine and, ultimately, affect every patient.
Pre-emptive sequencing of patients interrogates large numbers of pharmacogenomic variants and integrates clinically actionable results into a patient’s electronic health record (EHR) for efficient use by clinicians at the point of care. It is unknown, however, whether integration of pre-emptive pharmacogenomic data into the EHR will significantly improve patient outcomes and reduce health care costs. Therefore, the Right Drug, Right Dose, Right Time: Using Genomic Data to Individualize Treatment Protocol (RIGHT Protocol) study was designed to recruit a large group of patients for pre-emptive pharmacogenomic testing, to develop the EHR infrastructure to deliver point of care clinical decision support and to study the effects of integrating pre-emptive pharmacogenomic testing into applied clinical practice on patient outcomes. As a result of these efforts, we have recruited 11 098 patients to study pre-emptive pharmacogenomic implementation in an integrated medical system. The RIGHT cohort therefore serves as a scientific resource to broadly address the following scientific questions.
What are the associations between genetic variants and clinical outcomes in patients prescribed drugs known to interact with these genes? The outcomes we will investigate include therapeutic response (e.g. pain control), time to optimal therapeutic goal, number of drug and/or dose changes, health care utilization (e.g. clinic visits and hospital days), adverse drug events/toxicity (e.g. gastro-intestinal upset), and subsequent clinical events (e.g. stroke).
Does pre-emptive pharmacogenomic testing improve patient outcomes and reduce health care costs through optimized treatment? By comparing patients with pharmacogenomic data to pharmacogenomics-naïve patients we will test the hypothesis that pre-emptive pharmacogenomic testing reduces adverse drug events, lowers health care utilizations and costs, and improves therapeutic response, patient safety and satisfaction.
What clinical outcomes are associated with pharmacogenomic variants of unknown or indeterminate clinical significance or for which evidence for changes in prescribing practices is limited? A critical barrier in identifying pharmacogenomic variants that are clinically significant is the reliance on costly clinical trials that investigate a single drug or outcome. Observational studies such as the RIGHT Protocol study will provide valuable insights regarding the cost-effectiveness of identifying drug–gene interactions that result in adverse drug events and other clinically significant outcomes.
How does pre-emptive pharmacogenomic testing impact clinician and patient perspectives and the broad spectrum of clinical practice? Understanding the potential influence of pre-emptive pharmacogenomics in prescribing practices, as well as identifying the possible impacts to clinician and patient interactions, can elucidate risks and opportunities that support safe integration. Identifying gaps and providing education is vital to support wide-scale implementation when diffusing medical innovations.
Who is in the cohort?
Setting
Mayo Clinic is a non-profit academic medical institution with major campuses in Rochester, MN; Scottsdale and Phoenix, AZ; and Jacksonville, FL. The Mayo Clinic Health Systems has dozens of locations in MN, IA and WI. Together, these sites provide comprehensive patient care, education in clinical medicine and medical sciences, and extensive programmes in research. In 2018, Mayo Clinic provided direct care for 1.3 million people from all 50 states and 136 countries.
The Mayo Clinic Biobank is an institutional resource comprised of adult community volunteers who donated biological specimens, provided risk factor data, allowed access to EHR data for research and consented to participate in additional studies.7 To date, over 57 000 subjects have consented to participate. Since its inception, this resource has been used for more than 250 ancillary studies and has distributed over 125 000 samples. The Mayo Clinic Biobank was the source cohort for the RIGHT Protocol study, which allowed the timely recruitment of 11 098 participants with previously stored DNA samples.
Mayo Clinic EHR data are available for all Biobank participants. However, patients use different health care providers for different reasons and may change providers and health care systems over time. Therefore, the Mayo Clinic EHR may not contain complete data for all Biobank participants. However, most of the Biobank participants (82%) reside within the region captured by the Rochester Epidemiology Project (REP) medical records-linkage system. The REP is an established research infrastructure that links the health care records of persons residing in a 27-county region of southern MN and western WI (approximately 1.1 million persons in 2010).8–10 Briefly, the REP links the health care records of Mayo Clinic, Olmsted Medical Center and its satellites, Olmsted County Public Health Services, Zumbro Valley Health Center and some smaller health care providers in this region. As residents of this region visit one of the REP collaborators for health care, the REP captures the electronic data related to the health care visits. In addition, the full texts of health care records are available for abstraction for all persons in the REP system since 1966. Overall, 99% of the Biobank participants have at least one non-Mayo Clinic medical record captured by the REP. Therefore, linking the Biobank participants to the REP infrastructure provides access to more complete coverage and follow-up of the Biobank population.
Recruitment
Beginning in 2012, Biobank participants were recruited for the RIGHT pilot study, which has been previously described.11 In brief, 1013 patients consented to pharmacogenomic testing and return of pharmacogenomic results to their EHR for use in their clinical care. This pilot resulted in the infrastructure to support the expansion of the cohort. Beginning in May 2016, we began a targeted effort to further enroll an additional 10 000 persons. Patients with prior allogeneic haematopoietic stem cell transplants and/or haematologic malignancies were not eligible to participate due to the fact that circulating lymphocytes were used as the DNA source. All non-White Biobank participants (n = 1442) were preferentially invited to increase the diversity of the cohort. Participants were then prioritized by those that receive the majority of their care at Mayo Clinic based on EHR length and a history of clinic visits in a primary care clinic (e.g. Internal Medicine, Family Medicine). All eligible Biobank participants were mailed an invitation to participate in the RIGHT Protocol study and a consent form. The invitation letter gave recipients three options: (i) participate in the study, (ii) decline to participate, or (iii) declare themselves ineligible because they did not intend to utilize Mayo Clinic for their future health care. Consenting RIGHT participants agreed to allow use of their stored Biobank samples for clinical pharmacogenomic testing, deposit of pharmacogenomic results into their EHR for clinical use, and use of EHR and pharmacogenomic data for research. If no response was received after a period of 4 weeks, one additional attempt was made with a second mailing. This expanded recruitment effort was completed in 16 months. Overall, 18 199 persons were approached, 56% consented to participate, 29% did not respond and 15% self-reported ineligibility or refused (Table 1).
Table 1.
Characteristics of 1013 RIGHT pilot participants and the 18 199 patients invited to participate in the expansion of the RIGHT protocol
| Characteristic | RIGHT expansion results |
|||||
|---|---|---|---|---|---|---|
| RIGHT pilot participantsa | Invited | Consented | Did not respond | Ineligible/ refused | Complete RIGHT cohort | |
| n = 1013 | n = 18 199 | n = 10 085 | n = 5331 | n = 2783 | n = 11 098 | |
| Sex, n (%) | ||||||
| Female | 538 (53) | 11 043 (61) | 6150 (61) | 3192 (60) | 1701 (61) | 6688 (60) |
| Male | 475 (47) | 7156 (39) | 3935 (39) | 2139 (40) | 1082 (39) | 4410 (40) |
| Age on January 1, 2016, years, n (%) | ||||||
| 18–24 | 0 (0) | 151 (1) | 58 (1) | 76 (1) | 17 (1) | 58 (1) |
| 25–34 | 0 (0) | 1701 (9) | 647 (6) | 784 (15) | 270 (10) | 647 (6) |
| 35–44 | 2 (<1) | 1722 (9) | 825 (8) | 744 (14) | 153 (6) | 827 (7) |
| 45–54 | 178 (18) | 2390 (13) | 1299 (13) | 831 (16) | 260 (9) | 1477 (13) |
| 55–64 | 658 (65) | 3769 (21) | 2069 (21) | 1138 (21) | 562 (20) | 2727 (25) |
| 65–74 | 175 (17) | 5084 (28) | 3217 (32) | 1056 (20) | 811 (29) | 3392 (31) |
| 75+ | 0 (0) | 3382 (19) | 1970 (20) | 702 (13) | 710 (26) | 1970 (18) |
| Race, n (%) | ||||||
| White | 966 (95) | 16 759 (92) | 9482 (94) | 4689 (88) | 2588 (93) | 10 448 (94) |
| Non-White | 39 (4) | 1177 (6) | 524 (5) | 492 (9) | 161 (6) | 563 (5) |
| Black | 7 (1) | 124 (1) | 50 (<1) | 64 (1) | 10 (<1) | 57 (1) |
| Asian | 9 (1) | 276 (2) | 91 (1) | 139 (3) | 46 (2) | 100 (1) |
| AIAN | 0 (0) | 30 (<1) | 16 (<1) | 8 (<1) | 6 (<1) | 16 (<1) |
| NHPI | 0 (0) | 0 (0) | 0 (0) | 0 (0) | 0 (0) | 0 (0) |
| Other and mixed | 23 (2) | 747 (4) | 367 (4) | 281 (5) | 99 (4) | 390 (3) |
| Unknown | 8 (1) | 263 (1) | 79 (1) | 150 (3) | 34 (1) | 87 (1) |
| Ethnicity, n (%) | ||||||
| Non-Hispanic | 1006 (99) | 17 906 (98) | 9967 (99) | 5205 (98) | 2734 (98) | 10 973 (99) |
| Hispanic | 7 (1) | 275 (2) | 112 (1) | 115 (2) | 48 (2) | 119 (1) |
| Unknown | 0 (0) | 18 (<1) | 6 (<1) | 11 (<1) | 1 (<1) | 6 (<1) |
| Self-reported education at time of Biobank consent (2009–2017), n (%) | ||||||
| High school graduate or General Educational Development (GED) or less | 87 (9) | 2754 (15) | 1262 (13) | 897 (17) | 595 (21) | 1349 (12) |
| Some college or Associates degree (including community college) | 326 (32) | 5614 (31) | 2938 (29) | 1817 (34) | 859 (31) | 3264 (29) |
| Four year college graduate (Bachelor's degree) | 238 (24) | 3580 (20) | 1992 (20) | 1079 (20) | 509 (18) | 2230 (20) |
| Graduate or professional school | 362 (36) | 6116 (34) | 3848 (38) | 1476 (28) | 792 (28) | 4210 (38) |
| Unknown | 0 (0) | 135 (1) | 45 (<1) | 62 (1) | 28 (1) | 45 (<1) |
Previously recruited in 2012–2013.
AIAN, American Indian or Alaska Native; NHPI, Native Hawaiian or Pacific Islander.
Cohort characteristics
Table 1 summarizes the characteristics of the RIGHT pilot participants, recruitment results for the expansion of the cohort and the characteristics of the full RIGHT cohort. Table 2 compares characteristics of the RIGHT cohort, Mayo Clinic Biobank and REP 27-county populations.
Table 2.
Characteristics as of January 1, 2016 of the RIGHT cohort, Mayo Clinic Biobank and Rochester Epidemiology Project
| RIGHT cohort | Mayo Clinic Biobank | Rochester Epidemiology Project 27-counties | |
|---|---|---|---|
| Characteristic | n = 11 098 | n = 56 988 | n = 582 466 |
| Sex, n (%) | |||
| Female | 6688 (60) | 33 478 (59) | 309 707 (53) |
| Male | 4410 (40) | 23 510 (41) | 272 759 (47) |
| Age on January 1, 2016, years, n (%) | |||
| 18–24 | 58 (1) | 536 (1) | 69 237 (12) |
| 25–34 | 647 (6) | 3227 (6) | 98 800 (17) |
| 35–44 | 827 (7) | 4026 (7) | 85 385 (15) |
| 45–54 | 1477 (13) | 7238 (13) | 94 059 (16) |
| 55–64 | 2727 (25) | 13 020 (23) | 101 644 (17) |
| 65–74 | 3392 (31) | 15 115 (27) | 71 216 (12) |
| 75+ | 1970 (18) | 13 826 (24) | 62 125 (11) |
| Race, n (%) | |||
| White | 10 448 (94) | 54 316 (95) | 520 716 (89) |
| Non-White | 563 (5) | 1967 (3) | 41 434 (7) |
| Black | 57 (1) | 702 (1) | 14 248 (2) |
| Asian | 100 (1) | 572 (1) | 11 423 (2) |
| AIAN | 16 (<1) | 117 (<1) | 1491 (<1) |
| NHPI | 0 (0) | 22 (<1) | 842 (<1) |
| Other and mixed | 390 (3) | 554 (1) | 13 430 (2) |
| Unknown race | 87 (1) | 705 (1) | 20 316 (3) |
| Ethnicity, n (%) | |||
| Non-Hispanic | 10 973 (99) | 5370 (94) | 505 374 (87) |
| Hispanic | 119 (1) | 783 (1) | 22 428 (4) |
| Unknown | 6 (<1) | 2504 (4) | 54 664 (9) |
| Socio-economic characteristics | |||
| ≥ High school (%) | 99 | 97 | 41* |
| ≥ Bachelor’s degree (%) | 58 | 50 | 15* |
Education data are available for 329 589 (57%) persons; percentages are based on this denominator.
AIAN, American Indian or Alaska Native; NHPI, Native Hawaiian, or Pacific Islander.
How often have they been followed up?
Prospective and retrospective studies
A primary goal of the RIGHT Protocol study was to implement pre-emptive pharmacogenomics for Mayo Clinic patients. Therefore, by design, we anticipate the majority of our follow-up will be conducted passively, via the EHR, in order to estimate patient outcomes after implementation of pre-emptive pharmacogenomics. Prospective studies, defined as occurring after pharmacogenomic data were placed in the EHR, permit comparisons of the RIGHT cohort to a pharmacogenomics-naïve group to determine the effects of genotype-guided prescribing. As of January 1, 2018, 99% of the cohort was living and 85% had a clinic visit in 2017 (Table 3). The RIGHT Protocol study also enables retrospective studies by leveraging decades of drug exposures to compare drug response across genotype prior to pharmacogenomic results being placed in the EHR (pharmacogenomics-naïve time period). The median length of the medical record is 22 years for the cohort (Table 3). For either study type, outcomes can include health care utilization, drug type and dose changes, time to optimal therapy, and adverse drug events including subsequent clinical events (e.g. stent thrombosis). Finally, the availability of genetic sequence data allows for discovery of additional drug–gene interactions for which clinical evidence is limited or non-existent.
Table 3.
Medical record data for the RIGHT cohort
| Characteristics | RIGHT cohort |
|---|---|
| n = 11 098 | |
| Alive as of January 1, 2018, n (%) | 11 021 (99) |
| Clinical contact in 2017, % yes | 85 |
| Length of medical record prior to 2017, median years interquartile range (IQR) | 22 (19, 27) |
| Clinic visits per year from 2004 to 2017, median visits (IQR) | 27 (16, 44) |
| Prescribed medications in 2017, n (%) | |
| None | 1010 (9) |
| 1–2 | 1725 (16) |
| 3–5 | 2732 (25) |
| 6–10 | 2992 (27) |
| 11–20 | 2124 (19) |
| ≥ 21 | 515 (5) |
| Prescribed a drug with therapeutic recommendationsa 2004–2017, n (%) | |
| abacavir | 2 (<1) |
| allopurinol | 494 (4) |
| amitriptyline | 664 (6) |
| atazanavir | 1 (<1) |
| azathioprine | 134 (1) |
| capecitabine | 20 (<1) |
| carbamazepine | 102 (1) |
| citalopram | 1531 (14) |
| escitalopram | 524 (5) |
| clomipramine | 8 (<1) |
| clopidogrel | 636 (6) |
| codeine | 3704 (33) |
| desipramine | 20 (<1) |
| doxepin | 179 (2) |
| fluorouracil | 625 (6) |
| fluvoxamine | 16 (<1) |
| imipramine | 29 (<1) |
| ivacaftor | 0 (0) |
| mercaptopurine | 11 (<1) |
| nortriptyline | 723 (7) |
| ondansetron | 1927 (17) |
| oxcarbazepine | 29 (<1) |
| paroxetine | 369 (3) |
| peginterferon alfa-2a | 4 (<1) |
| peginterferon alfa-2b | 3 (<1) |
| ribavirin | 8 (<1) |
| phenytoin | 41 (<1) |
| rasburicase | 0 (0) |
| sertraline | 1268 (11) |
| simvastatin | 3562 (32) |
| tacrolimus | 449 (4) |
| tamoxifen | 180 (2) |
| tegafur | 0 (0) |
| thioguanine | 0 (0) |
| trimipramine | 3 (<1) |
| tropisetron | 0 (0) |
| voriconazole | 53 (<1) |
Therapeutic recommendations are available at https://www.pharmgkb.org/guidelines
Active follow-up
Participants in the cohort may, with appropriate approvals, be contacted directly to assess pharmacogenomics-related outcomes. To date, we have conducted three surveys of RIGHT participants. The RIGHT pilot study participants were surveyed in 2014 to assess their understanding of their CYP2D6 results.12 Of those invited, 86% completed the survey.13 In 2017, 5000 subjects from the expanded cohort were surveyed after enrollment to collect information on prior testing, expectations and concerns about pharmacogenomic testing, as well as their views about their current medications and medications in general. Response to this survey was exceptionally high (94%) and reflects a very high level of interest in pharmacogenomics in our population. A follow-up survey is currently in the field to assess responses following receipt of pharmacogenomic results. Additionally, there are opportunities to study clinician perspectives, pharmacogenomic testing behaviours and clinical actions following deposition of pharmacogenomic results into the EHR.
Loss to follow-up
The recruitment design of the study prioritized patients for whom Mayo Clinic is their primary care clinic and therefore loss to follow-up is mitigated. Likewise, use of the REP allows for data capture within the surrounding community practices. However, data capture decreases as the distance of the patient’s residence from Rochester, MN increases. Loss to follow-up will occur for those patients moving or choosing a provider out of our catchment area. However, as indicated in Table 3, the RIGHT cohort reflects a stable patient population with 91% receiving at least one prescription in 2017.
What has been measured?
DNA sequencing and pharmacogenomic phenotyping
For the 1013 pilot-study participants, lymphocyte-derived DNA sequencing of 84 pharmacogenes was completed at Mayo Clinic using Version 1 (v.1) of the PGRN-Seq assay.14 For the expanded RIGHT cohort, the Clinical Laboratory Improvement Amendments (CLIA)-certified and College of American Pathologists (CAP)-accredited Baylor College of Medicine’s Human Genome Sequencing Center Clinical Laboratory sequenced 77 genes using version 3 (v.3.) of the PGRN-Seq assay (now termed PGx-seq). This panel includes structurally complex genomic regions important to optimal drug treatment including CYP2D6 and HLA regions. CYP2D6 is involved in the metabolism of 25% of the drugs currently on the market15 and HLA regions are associated with severe drug hypersensitivities (e.g. Stevens-Johnson Syndrome). In v.1, all gene coverage included all gene model exons plus both 2 kb upstream and 1 kb downstream. In v.1, targeted exons and 2 kb upstream/1 kb downstream for CYP2A6 were included. For PGx-seq, these upstream and downstream targets were dropped. In both v.1 and PGx-seq, intron 6 of CYP3A4 was targeted. In PGx-seq, the entire ∼31 kb region for CYP2D6 was targeted including both of the nearby pseudogenes. For both versions, all of the Illumina ADME and Affymetrix DMET single nucleotide polymorphism (SNP) sites not already covered by gene targets were included. Over 250 genes have either DNA sequencing or SNP genotyping information available (Table 4).
Table 4.
Genetic data available in the RIGHT study
| Gene | Sequence data |
Genotype data | |
|---|---|---|---|
| Version 1 | Version 3 | ||
| ABCA1 | Yes | No | No |
| ABCB1 | Yes | Yes | Yes |
| ABCB11 | Yes | No | Yes |
| ABCB4 | No | No | Yes |
| ABCB7 | No | No | Yes |
| ABCC1 | No | No | Yes |
| ABCC2 | Yes | No | Yes |
| ABCC3 | No | No | Yes |
| ABCC4 | No | Yes | Yes |
| ABCC5 | No | No | Yes |
| ABCC6 | No | No | Yes |
| ABCC8 | No | No | Yes |
| ABCC9 | No | No | Yes |
| ABCG1 | Yes | No | Yes |
| ABCG2 | Yes | Yes | Yes |
| ABP1 | No | No | Yes |
| ACAD10 | No | No | Yes |
| ACE | Yes | No | No |
| ADH1A | No | No | Yes |
| ADH1B | No | No | Yes |
| ADH1C | No | No | Yes |
| ADH4 | No | No | Yes |
| ADH5 | No | No | Yes |
| ADH6 | No | No | Yes |
| ADH7 | No | No | Yes |
| ADRB1 | Yes | No | No |
| ADRB2 | Yes | Yes | No |
| AGR3 | No | No | Yes |
| AHR | Yes | No | Yes |
| AKAP9 | No | No | Yes |
| ALB | No | No | Yes |
| ALDH1A1 | No | No | Yes |
| ALDH2 | No | No | Yes |
| ALDH3A1 | No | No | Yes |
| ALDH3A2 | No | No | Yes |
| ALOX5 | Yes | No | No |
| ANKK1 | No | Yes | No |
| ANKRD17 | No | No | Yes |
| ANXA1 | No | No | Yes |
| AOX1 | No | No | Yes |
| AOX2P | No | No | Yes |
| APOA1 | Yes | No | No |
| APOA2 | No | No | Yes |
| ARID5B | Yes | No | No |
| ARNT | No | No | Yes |
| ARSA | No | No | Yes |
| ASB4 | No | No | Yes |
| ATP7A | No | No | Yes |
| ATP7B | No | No | Yes |
| ATXN7L2 | No | No | Yes |
| BCKDK | No | No | Yes |
| BDNF | Yes | No | No |
| C14orf164 | No | No | Yes |
| C16orf54 | No | No | Yes |
| C18orf16 | No | No | Yes |
| C1orf87 | No | No | Yes |
| C3orf22 | No | No | Yes |
| C4orf17 | No | No | Yes |
| CACNA1C | Yes | No | No |
| CACNA1S | Yes | No | No |
| CACNB2 | Yes | No | No |
| CBR1 | No | No | Yes |
| CBR3 | No | No | Yes |
| CCDC101 | No | No | Yes |
| CDA | No | No | Yes |
| CES1 | Yes | Yes | No |
| CES2 | Yes | No | Yes |
| CFTR | No | Yes | No |
| CHST1 | No | No | Yes |
| CHST10 | No | No | Yes |
| CHST11 | No | No | Yes |
| CHST13 | No | No | Yes |
| CHST2 | No | No | Yes |
| CHST3 | No | No | Yes |
| CHST4 | No | No | Yes |
| CHST5 | No | No | Yes |
| CHST6 | No | No | Yes |
| CHST7 | No | No | Yes |
| CHST8 | No | No | Yes |
| CHST9 | No | No | Yes |
| COMT | Yes | Yes | Yes |
| CRHR1 | Yes | No | No |
| CRIP3 | No | No | Yes |
| CROT | No | No | Yes |
| CYP11A1 | No | No | Yes |
| CYP11B1 | No | No | Yes |
| CYP11B2 | No | No | Yes |
| CYP17A1 | No | No | Yes |
| CYP19A1 | No | No | Yes |
| CYP1A1 | No | No | Yes |
| CYP1A2 | Yes | Yes | Yes |
| CYP1B1 | No | No | Yes |
| CYP20A1 | No | No | Yes |
| CYP21A2 | No | No | Yes |
| CYP24A1 | No | No | Yes |
| CYP26A1 | No | No | Yes |
| CYP26C1 | No | No | Yes |
| CYP27A1 | No | No | Yes |
| CYP27B1 | No | No | Yes |
| CYP2A13 | No | No | Yes |
| CYP2A6 | Yes | Yes | Yes |
| CYP2A7 | No | No | Yes |
| CYP2B6 | Yes | Yes | Yes |
| CYP2B7P1 | No | No | Yes |
| CYP2C18 | No | No | Yes |
| CYP2C19 | Yes | Yes | Yes |
| CYP2C8 | No | Yes | Yes |
| CYP2C9 | Yes | Yes | Yes |
| CYP2D6 | Yes | Yes | Yes |
| CYP2D7P1 | No | No | Yes |
| CYP2E1 | No | Yes | Yes |
| CYP2F1 | No | No | Yes |
| CYP2J2 | No | Yes | Yes |
| CYP2R1 | Yes | No | No |
| CYP2S1 | No | No | Yes |
| CYP39A1 | No | No | Yes |
| CYP3A4 | Yes | Yes | Yes |
| CYP3A43 | No | No | Yes |
| CYP3A5 | Yes | Yes | Yes |
| CYP3A7 | No | No | Yes |
| CYP46A1 | No | No | Yes |
| CYP4A11 | No | No | Yes |
| CYP4A22 | No | No | Yes |
| CYP4B1 | No | No | Yes |
| CYP4F11 | No | No | Yes |
| CYP4F12 | No | No | Yes |
| CYP4F2 | No | Yes | Yes |
| CYP4F3 | No | No | Yes |
| CYP4F8 | No | No | Yes |
| CYP4X1 | No | No | Yes |
| CYP4Z1 | No | No | Yes |
| CYP4Z2P | No | No | Yes |
| CYP51A1 | No | No | Yes |
| CYP7A1 | No | No | Yes |
| CYP7B1 | No | No | Yes |
| CYP8B1 | No | No | Yes |
| DBH | Yes | No | No |
| DCK | No | No | Yes |
| DKFZp779M0652 | No | No | Yes |
| DPYD | Yes | Yes | Yes |
| DRD1 | Yes | No | No |
| DRD2 | Yes | Yes | No |
| DRD3 | No | Yes | No |
| DRD4 | No | Yes | No |
| EGFR | Yes | Yes | No |
| EPHX1 | No | No | Yes |
| EPHX2 | No | No | Yes |
| ESR1 | Yes | No | No |
| F2 | No | No | No |
| F5 | No | Yes | No |
| FAAH | No | Yes | Yes |
| FAM19A3 | No | No | Yes |
| FKBP5 | Yes | No | No |
| FMO1 | No | No | Yes |
| FMO2 | No | No | Yes |
| FMO3 | No | No | Yes |
| FMO4 | No | No | Yes |
| FMO5 | No | No | Yes |
| FMO6P | No | No | Yes |
| G6PD | Yes | Yes | Yes |
| GGCX | No | Yes | No |
| GLCCI1 | Yes | No | No |
| GRIK4 | No | Yes | No |
| GRIP2 | No | No | Yes |
| GRK4 | Yes | No | No |
| GRK5 | Yes | No | No |
| GSTA1 | No | No | Yes |
| GSTA2 | No | No | Yes |
| GSTA3 | No | No | Yes |
| GSTA4 | No | No | Yes |
| GSTA5 | No | No | Yes |
| GSTM1 | No | No | Yes |
| GSTM2 | No | No | Yes |
| GSTM2P1 | No | No | Yes |
| GSTM3 | No | No | Yes |
| GSTM4 | No | No | Yes |
| GSTM5 | No | No | Yes |
| GSTO1 | No | No | Yes |
| GSTP1 | No | Yes | Yes |
| GSTT1 | No | No | Yes |
| GSTT2 | No | No | Yes |
| GSTT2B | No | No | Yes |
| GSTZ1 | No | No | Yes |
| HLA-A | No | Yes | No |
| HLA-B | No | Yes | No |
| HMGCR | Yes | Yes | Yes |
| HNF1A | No | Yes | No |
| HNF4A | No | Yes | No |
| HNMT | No | No | Yes |
| HOMEZ | No | No | Yes |
| HSD11B2 | Yes | No | No |
| HTR1A | Yes | No | No |
| HTR2A | Yes | Yes | No |
| HTR2C | No | Yes | No |
| ICK | No | No | Yes |
| IFNL3 | No | Yes | No |
| IGFBP7 | No | Yes | No |
| IL28B | No | No | No |
| KCNH2 | Yes | Yes | No |
| KIF6 | No | Yes | No |
| LDLR | Yes | Yes | No |
| LEP | No | Yes | No |
| LEPR | No | Yes | No |
| LOC26172 | No | No | Yes |
| LOC286186 | No | No | Yes |
| LOC619207 | No | No | Yes |
| LOC651536 | No | No | Yes |
| LOC96610 | No | No | Yes |
| LST-3TM12 | No | No | Yes |
| MAOA | Yes | No | Yes |
| MAOB | No | No | Yes |
| MAT1A | No | No | Yes |
| METTL1 | No | No | Yes |
| MKL1 | No | No | Yes |
| MTHFR | No | Yes | No |
| MYOF | No | No | Yes |
| NAT1 | No | No | Yes |
| NAT2 | Yes | Yes | Yes |
| NNMT | No | No | Yes |
| NPPB | Yes | No | No |
| NPR1 | Yes | No | No |
| NQO1 | No | No | Yes |
| NR1I2 | No | No | Yes |
| NR1I3 | No | No | Yes |
| NR3C1 | Yes | No | Yes |
| NR3C2 | Yes | No | No |
| NTRK2 | Yes | No | No |
| NTSR1 | No | No | Yes |
| NUDT15 | No | No | No |
| OPA3 | No | No | Yes |
| OPRM1 | No | Yes | No |
| ORM1 | No | No | Yes |
| ORM2 | No | No | Yes |
| PARL | No | No | Yes |
| PCP4L1 | No | No | Yes |
| PEAR1 | Yes | No | No |
| PGAP3 | No | No | Yes |
| PNMT | No | No | Yes |
| PON1 | No | Yes | Yes |
| PON2 | No | No | Yes |
| PON3 | No | No | Yes |
| POR | Yes | No | Yes |
| PPARD | No | No | Yes |
| PPARG | No | No | Yes |
| PPP1R9A | No | No | Yes |
| PRSS53 | No | No | Yes |
| PTGIS | Yes | No | Yes |
| PTGS1 | Yes | No | No |
| QPRT | No | No | Yes |
| RALBP1 | No | No | Yes |
| RPL13 | No | No | Yes |
| RXRA | No | No | Yes |
| RYR1 | Yes | Yes | No |
| RYR2 | Yes | Yes | No |
| SCN1A | No | Yes | No |
| SCN5A | Yes | Yes | No |
| SDCBP | No | No | Yes |
| SERPINA7 | No | No | Yes |
| SETD4 | No | No | Yes |
| SHANK3 | No | No | Yes |
| SLC10A1 | No | No | Yes |
| SLC10A2 | No | No | Yes |
| SLC13A1 | No | No | Yes |
| SLC15A1 | No | No | Yes |
| SLC15A2 | Yes | No | Yes |
| SLC16A1 | No | No | Yes |
| SLC19A1 | No | Yes | Yes |
| SLC22A1 | Yes | Yes | Yes |
| SLC22A11 | No | No | Yes |
| SLC22A12 | No | No | Yes |
| SLC22A13 | No | No | Yes |
| SLC22A14 | No | No | Yes |
| SLC22A2 | Yes | Yes | Yes |
| SLC22A3 | Yes | No | Yes |
| SLC22A4 | No | No | Yes |
| SLC22A5 | No | No | Yes |
| SLC22A6 | Yes | No | Yes |
| SLC22A7 | No | No | Yes |
| SLC22A8 | No | No | Yes |
| SLC25A27 | No | No | Yes |
| SLC28A1 | No | No | Yes |
| SLC28A2 | No | No | Yes |
| SLC28A3 | No | No | Yes |
| SLC29A1 | No | No | Yes |
| SLC29A2 | No | No | Yes |
| SLC41A2 | No | No | Yes |
| SLC47A1 | Yes | No | No |
| SLC47A2 | Yes | No | No |
| SLC5A6 | No | No | Yes |
| SLC6A3 | Yes | No | No |
| SLC6A4 | Yes | Yes | No |
| SLC6A6 | No | No | Yes |
| SLC7A5 | No | No | Yes |
| SLC7A5P1 | No | No | Yes |
| SLC7A7 | No | No | Yes |
| SLC7A8 | No | No | Yes |
| SLC9A7 | No | No | Yes |
| SLCO1A2 | Yes | No | Yes |
| SLCO1B1 | Yes | Yes | Yes |
| SLCO1B3 | Yes | No | Yes |
| SLCO2B1 | Yes | Yes | Yes |
| SLCO3A1 | No | No | Yes |
| SLCO4A1 | No | No | Yes |
| SLCO5A1 | No | No | Yes |
| SOD2 | No | Yes | No |
| SPG7 | No | No | Yes |
| SPN | No | No | Yes |
| SR140 | No | No | Yes |
| SULT1A1 | No | Yes | Yes |
| SULT1A2 | No | No | Yes |
| SULT1A3 | No | No | Yes |
| SULT1A4 | No | No | Yes |
| SULT1B1 | No | No | Yes |
| SULT1C2 | No | No | Yes |
| SULT1C4 | No | No | Yes |
| SULT1E1 | No | No | Yes |
| SULT2A1 | No | No | Yes |
| SULT2B1 | No | No | Yes |
| SULT4A1 | No | No | Yes |
| TBXAS1 | Yes | No | Yes |
| TCL1A | Yes | No | No |
| TFF3 | No | No | Yes |
| TMEM120A | No | No | Yes |
| TMEM176A | No | No | Yes |
| TMPRSS11B | No | No | Yes |
| TOP1P1 | No | No | Yes |
| TP53 | No | Yes | No |
| TPMT | Yes | Yes | Yes |
| TPSG1 | No | No | Yes |
| TSEN2 | No | No | Yes |
| TTBK1 | No | No | Yes |
| TYMS | No | Yes | Yes |
| UGT1A1 | Yes | Yes | Yes |
| UGT1A10 | No | No | Yes |
| UGT1A3 | No | No | Yes |
| UGT1A3-10 exon1 | No | Yes | No |
| UGT1A4 | Yes | Yes | Yes |
| UGT1A5 | No | No | Yes |
| UGT1A6 | No | No | Yes |
| UGT1A7 | No | No | Yes |
| UGT1A8 | No | No | Yes |
| UGT1A9 | No | No | Yes |
| UGT2A1 | No | No | Yes |
| UGT2A2 | No | No | Yes |
| UGT2B10 | No | No | Yes |
| UGT2B11 | No | No | Yes |
| UGT2B15 | No | Yes | Yes |
| UGT2B17 | No | No | Yes |
| UGT2B28 | No | No | Yes |
| UGT2B4 | No | No | Yes |
| UGT2B7 | No | Yes | Yes |
| UGT8 | No | No | Yes |
| VDR | Yes | No | No |
| VEGFA | No | Yes | No |
| VKORC1 | Yes | Yes | Yes |
| XDH | No | No | Yes |
| XYLB | No | No | Yes |
| ZNF280B | No | No | Yes |
| ZNF423 | Yes | No | No |
| ZNF662 | No | No | Yes |
For the RIGHT pilot, CYP2C19, CYP2C9, VKORC1, SLCO1B1 and CYP2D6, sequencing results were interpreted at the Mayo Clinic Personalized Genomics Laboratory, a CLIA-certified and CAP-accredited laboratory. These genes were chosen because they have strong clinical practice recommendations when patients are prescribed warfarin, clopidogrel, simvastatin, tamoxifen, opioid analgesics and selective serotonin reuptake inhibitors.16 For the expanded RIGHT cohort, CYP1A2, CYP2C9, CYP2C19, CYP2D6, CYP3A4, CYP3A5, DPYD, SLCO1B1, TPMT, UGT1A1, VKORC1, HLA-A and HLA-B are interpreted and reported in the EHR by the Personalized Genomics Laboratory. Point of care drug–gene alerts access these reports and present a description of the variants identified for each gene to the prescriber. In addition, the clinician receives an interpretation of the impact of the findings on gene function as well as pharmacogenomic implications. Table 5 includes the drug–gene alerts active in the Mayo Clinic EHR as of December 31, 2018.
Table 5.
Mayo Clinic clinical decision support interventions
| Gene | Drug | Year went live |
|---|---|---|
| CYP2C19 | Clopidogrel | 2014 |
| Citalopram | 2015 | |
| Escitalopram | 2015 | |
| CYP2D6 | Codeine | 2013 |
| Tramadol | 2013 | |
| Tamoxifen | 2013 | |
| Paroxetine | 2015 | |
| Fluoxetine | 2015 | |
| Fluvoxamine | 2015 | |
| Venlafaxine | 2015 | |
| SLCO1B1 | Simvastatin | 2014 |
| CYP2C9/VKORC1 | Warfarin | 2014 |
|
TPMT
Genotype or phenotype |
Mercaptopurine | 2013 |
| Thioguanine | 2013 | |
| Azathioprine | 2013 | |
| HLA-B*1502 | Carbamazepine | 2013 |
| HLA-B*5701 | Abacavir | 2013 |
| HLA-B*5801 | Allopurinol | 2014 |
| CYP3A5 | Tacrolimus | 2016 |
| DPYD | 5-FU | 2017 |
| NUDT15 | Mercaptopurine | 2018 |
| Thioguanine | 2018 | |
| Azathioprine | 2018 | |
| HLA-A*3101 | Carbamazepine | 2018 |
A pharmacogenomic interpretive report (OneOme LLC, Minneapolis, MN) provides information to clinicians on how an individual patient's genes may affect drug response. Proprietary algorithms and curated clinical data are used to generate a personalized report based on a patient's genomic results. The report lists drugs categorized into three different bins according to genomic impact on different metabolic or physiologic effects: ‘Major gene–drug interaction’ (red), ‘Moderate gene–drug interaction’ (yellow) and ‘Minimal gene–drug interaction’ (green). The drugs are further classified by primary therapeutic area. For each drug, the report provides the patient’s results of the interpreted genes along with additional clinical annotation and interpretations in both textual and graphical form to provide clinicians with a better understanding of the complex information. These static reports flow through the Personalized Genomics Laboratory and are placed into the EHR for each patient. These reports may be viewed by health care providers, but do not interact with the EHR.
Drug prescription information
Drug prescription information is available via the REP. Complete outpatient electronic prescription data are available from January 1, 2004 through to the present. Filled drug prescription data are not available. However, medication reconciliation occurs when patients visit a health care provider. Patients are asked to list all of the medications they are currently taking, including prescribed medications, over the counter medications and supplements. All medication information is coded using the RxNorm and National Drug File-Reference Terminology (NDF-RT) coding systems.17,18 Combination drugs with multiple ingredients are counted under the category of the main ingredient, or under the combination drug category when applicable. The REP electronic indexes may be searched to identify all persons with either a prescription or report of medication use in the desired time frame. Prescriptions for drugs with pharmacogenomic guidelines for the RIGHT cohort are summarized in Table 3.
Outcomes
Relevant pharmacogenomic outcomes (e.g. response to treatment or adverse events) may be assessed electronically using coded data from the available EHRs and billing systems. Such data include all International Classification of Diseases (ICD-9 and ICD-10) and Current Procedural Terminology (CPT) coded diagnoses and procedures from these EHRs, as well as health care utilizations including hospitalizations, office visits and emergency room visits. We also maintain access to the full text of the EHRs in the REP, and this text is available for chart abstraction or for natural language processing (NLP) techniques. We have used such techniques to identify prescription drug allergies from the linked EHRs (see below).
Processed allergy section clinical notes
Drug side-effects as well as drug allergies are commonly reported in the allergy section of the clinical note. As a result, the allergy section is a rich source of information for adverse drug reactions. However, descriptions of drug events are often embedded in both structured and unstructured clinical narratives, and thus processes for cleaning and standardizing the data are required. For the RIGHT Protocol study, we deployed NLP methods to process the clinical text and annotate clinical concepts to capture adverse drug events.19–21 We used the Medical Dictionary for Regulatory Activities (MedDRA),22,23 a medical terminology ontology adopted to describe adverse drug reactions and provide an internationally approved classification, and the Unified Medical Language System (UMLS, version 2012AB) (https://www.nlm.nih.gov/research/umls/),24 a knowledge resource for biomedical vocabularies developed by the US National Library of Medicine. Allergy data from the EHR were pulled and recorded as unstructured text, for example: ‘fluid retention, hypertension’ and ‘anxiety, distorted vision’. We used MedTagger,25 an NLP pipeline to extract adverse drug events mentioned in text. The extracted descriptions were then mapped to MedDRA terms and then aggregated using UMLS. For example, in the allergy text ‘fluid retention, hypertension’, ‘fluid retention’ was first extracted and mapped to MedDRA ‘body fluid retention’, and mapped to ‘Renal and urinary disorders’ and ‘Metabolism and nutrition disorders’, whereas ‘hypertension’ was extracted and mapped to ‘Vascular disorders’. The top five most common drug reactions were skin and subcutaneous tissue disorders, gastrointestinal disorders, immune system disorders, nervous system disorders, and general disorders and administration site conditions.
EHR integration and clinical decision support
Since 2013, Mayo Clinic has developed and implemented a comprehensive multidisciplinary model26 to integrate pharmacogenomic test results and clinical decision support (CDS) interventions in the Mayo Clinic unified EHR system (Table 5). Real-time pharmacogenomic-CDS alerts advise the need for pharmacogenomic testing when it is indicated, alert for potentially actionable drug–gene interactions, explain to prescribers the nature of the drug–gene interaction, and provide patient-specific therapeutic recommendations. The rules are embedded in the clinical workflow of all prescribers and pharmacists managing prescriptions (inpatient and outpatient prescription management systems). Multiple implementation and process metrics have been developed to allow post-implementation analyses. It is possible to measure when alerts are triggered and whether the clinician changes the initial prescription following the CDS alert. Such data allow for studies of clinician behaviours related to pharmacogenomics-CDS implementation.
What has it found? Key findings and publications
In the RIGHT pilot study, we found that 99% of participants had a clinically actionable variant when considering the five genes reported in the EHR (CYP2C19, CYP2C9, VKORC1, SLCO1B1 and CYP2D6).27 We have also reported our experience with integrating pharmacogenomic-CDS alerts into clinical practice in recent publications.16,26,28
Drug-focused studies include investigating drug response and clinical outcomes associated with exposure to opioids,29 diabetes treatment, proton pump inhibitors, statins, anaesthesia, antidepressants, anticoagulants, penicillin, beta blockers and antifungals. Other ancillary studies have assessed patient and provider perspectives on pharmacogenomics.12,13,28 Finally, additional studies have examined associations between pharmacogenomic phenotypes and health care utilization.30
One of our key findings is related to our first return of results to participants. After the pilot was completed, we returned CYP2D6 results to participants and surveyed them to determine their understanding of the data. Of the respondents, 26% said that they only somewhat understood their results and 7% did not understand them at all. Participants commonly suggested that results should be personalized, should refer to drugs they were currently taking, and should use layperson language.12
What are the main strengths and weaknesses?
The strengths of the RIGHT cohort derive from the large sample size, extensive EHR data, medical record access for abstraction, ability to study any drug exposure, and availability of sequence data for discovery. The RIGHT cohort are all Mayo Clinic Biobank participants who have existing, stored biospecimens (DNA, plasma, serum, white blood cells) and who can be re-contacted for future research questions or invited to participate in ancillary studies. The ability to cost-effectively re-contact participants expands the research capability of the cohort. Despite the strengths, the RIGHT cohort has some limitations. First, the cohort was not selected from patients with any particular diagnosis or drug indication. Thus, although the RIGHT study is one of the largest studies of pre-emptive pharmacogenomics, less frequent and rare drug exposures will be difficult to study in this population. Second, because this is an observational cohort, and not a clinical trial, patients in this study have been prescribed drugs based upon prescriber practices, the dictates of insurance coverage, and not randomized methods. Thus, inferences may be limited.
Can I get hold of the data? Where can I find out more?
The RIGHT cohort is a resource for pharmacogenomic research.11 As part of the infrastructure, a RIGHT Data Access Committee has been created to review data requests for use of RIGHT data. External access to the data is facilitated by the Mayo Clinic Biobank31 https://www.mayo.edu/research/centers-programs/mayo-clinic-biobank/overview
Profile in a nutshell
Pharmacogenomics is the hallmark of genomic medicine worldwide.
11 098 Mayo Clinic patients have consented to pre-emptive pharmacogenomic testing.
- The RIGHT study was developed to address several goals:
- identify the associations between genetic variants and clinical outcomes in patients prescribed drugs known to interact with these genes,
- determine if pre-emptive pharmacogenomic testing improves patient outcomes and reduces health care costs,
- identify clinical outcomes associated with pharmacogenomic variants of unknown or indeterminate clinical significance, and
- understand the perspectives and practices of clinicians and patients.
DNA sequencing or SNP genotyping information is available for over 250 genes.
Decades of prescription data are available for research via the Rochester Epidemiology Project (REP) medical records-linkage system.
Comparisons of the RIGHT cohort to a pharmacogenomic-naïve cohort allow for the study of the effects of genotype-guided dosing.
Funding
This work was supported by the Mayo Clinic Center for Individualized Medicine, the Robert D. and Patricia E. Kern Center for the Science of Health Care Delivery, and the National Institutes of Health grants [U19 GM61388 (The Pharmacogenomics Research Network), R01 GM28157, U01 HG005137, R01 GM125633, R01 AG034676 (The Rochester Epidemiology Project)], and {U01 HG06379 and U01 HG06379 Supplement [The Electronic Medical Record and Genomics (eMERGE) Network]}.
Conflict of interest: J.L.B. III, L.W., R.M.W., and the Mayo Clinic have stock ownership and have licensed intellectual property to OneOme LLC.
References
- 1. Weinshilboum R. Inheritance and drug response. N Engl J Med 2003;348:529–37. [DOI] [PubMed] [Google Scholar]
- 2. Weinshilboum RM, Wang L.. Pharmacogenomics: precision medicine and drug response. Mayo Clin Proc 2017;92:1711–22. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Wang L, McLeod HL, Weinshilboum RM.. Genomics and drug response. N Engl J Med 2011;364:1144–53. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Clinical Pharmacogenetics Implementation Consortium (CPIC®). https://cpicpgx.org/ (19 November 2018, date last accessed).
- 5.The Canadian Pharmacogenomics Network for Drug Safety (CPNDS). http://cpnds.ubc.ca/ (1 March 2019, date last accessed).
- 6.Dutch Pharmacogenetics Working Group (DPWG). The Dutch Pharmacogenetics Working Group (DPWG) was established in 2005 by the Royal Dutch Pharmacist's Association (KNMP). https://www.pharmgkb.org/page/dpwg (19 November 2018, date last accessed).
- 7. Olson JE, Ryu E, Johnson KJ. et al. The Mayo Clinic Biobank: a building block for individualized medicine. Mayo Clin Proc 2013;88:952–62. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Rocca WA, Grossardt BR, Brue SM. et al. Data Resource Profile: expansion of the Rochester Epidemiology Project medical records-linkage system (E-REP). Int J Epidemiol 2018;47:368–j. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. St Sauver JL, Grossardt BR, Yawn BP. et al. Data resource profile: the Rochester Epidemiology Project (REP) medical records-linkage system. Int J Epidemiol 2012;41:1614–24. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. St Sauver JL, Grossardt BR, Yawn BP, Melton LJ 3rd, Rocca WA.. Use of a medical records linkage system to enumerate a dynamic population over time: the Rochester Epidemiology Project. Am J Epidemiol 2011;173:1059–68. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Bielinski SJ, Olson JE, Pathak J. et al. Preemptive genotyping for personalized medicine: design of the Right Drug, Right Dose, Right Time-using Genomic Data to Individualize Treatment Protocol. Mayo Clin Proc 2014;89:25–33. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Olson JE, Rohrer Vitek C, Bell EJ. et al. Participant-perceived understanding and perspectives on pharmacogenomics: the Mayo Clinic RIGHT protocol (Right Drug, Right Dose, Right Time). Genet Med 2017;19:819–25. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Bielinski SJ, St Sauver JL, Olson JE. et al. Are patients willing to incur out-of-pocket costs for pharmacogenomic testing? Pharmacogenomics J 2017;17:1–3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Gordon AS, Fulton RS, Qin X, Mardis ER, Nickerson DA, Scherer S.. PGRNseq: a targeted capture sequencing panel for pharmacogenetic research and implementation. Pharmacogenet Genom 2016;26:161–68. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Zhou SF, Liu JP, Chowbay B.. Polymorphism of human cytochrome P450 enzymes and its clinical impact. Drug Metab Rev 2009;41:89–295. [DOI] [PubMed] [Google Scholar]
- 16. Caraballo PJ, Parkulo M, Blair D. et al. Clinical decision support to implement CYP2D6 drug-gene interaction. Stud Health Technol Inform 2015;216:946.. [PubMed] [Google Scholar]
- 17. Pathak J, Chute CG.. Analyzing categorical information in two publicly available drug terminologies: RxNorm and NDF-RT. J Am Med Inform Assoc 2010;17:432–39. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Pathak J, Murphy SP, Willaert BN. et al. Using RxNorm and NDF-RT to classify medication data extracted from electronic health records: experiences from the Rochester Epidemiology Project. AMIA Annual Symposium Proceedings AMIA Symposium 2011;2011:1089–98. [PMC free article] [PubMed] [Google Scholar]
- 19. Wang Y, Wang L, Rastegar-Mojarad M. et al. Clinical information extraction applications: a literature review. J Biomed Inform 2018;77:34–49. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Wang X, Hripcsak G, Markatou M, Friedman C.. Active computerized pharmacovigilance using natural language processing, statistics, and electronic health records: a feasibility study. J Am Med Inform Assoc 2009;16:328–37. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Wang X, Chase H, Markatou M, Hripcsak G, Friedman C.. Selecting information in electronic health records for knowledge acquisition. J Biomed Inform 2010;43:595–601. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Pearson RK, Hauben M, Goldsmith DI. et al. Influence of the MedDRA hierarchy on pharmacovigilance data mining results. Int J Med Inform 2009;78:e97–e103. [DOI] [PubMed] [Google Scholar]
- 23. Bousquet C, Lagier G, Lillo-Le Louët A, Le Beller C, Venot A, Jaulent MC. Appraisal of the MedDRA conceptual structure for describing and grouping adverse drug reactions. Drug Saf 2005;28:19–34. [DOI] [PubMed] [Google Scholar]
- 24. Bodenreider O. The Unified Medical Language System (UMLS): integrating biomedical terminology. Nucleic Acids Res 2004;32:D267–70. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Wagholikar K, Torii M, Jonnalagadda S, Liu H.. Feasibility of pooling annotated corpora for clinical concept extraction. AMIA Jt Summits Transl Sci Proc 2012;2012:38.. [PMC free article] [PubMed] [Google Scholar]
- 26. Caraballo PJ, Hodge LS, Bielinski SJ. et al. Multidisciplinary model to implement pharmacogenomics at the point of care. Genet Med 2017;19:421–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Ji Y, Skierka JM, Blommel JH. et al. Preemptive pharmacogenomic testing for precision medicine: a comprehensive analysis of five actionable pharmacogenomic genes using next-generation DNA sequencing and a customized CYP2D6 genotyping cascade. J Mol Diagn 2016;18:438–45. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. St Sauver JL, Bielinski SJ, Olson JE. et al. Integrating pharmacogenomics into clinical practice: promise vs reality. Am J Med 2016;129:1093–9.e1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. St Sauver JL, Olson JE, Roger VL. et al. CYP2D6 phenotypes are associated with adverse outcomes related to opioid medications. Pharmgenomics Pers Med 2017;10:217–27. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Takahashi PY, Ryu E, Pathak J. et al. Increased risk of hospitalization for ultrarapid metabolizers of cytochrome P450 2D6. Pharmgenomics Pers Med 2017;10:39–47. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Mayo Clinic Biobank. https://www.mayo.edu/research/centers-programs/mayo-clinic-biobank/overview (19 November 2018, date last accessed).