Secondary Logo

Journal Logo

Featured Articles: Editorial

Perioperative Precision Medicine and Bedside Decision Making: Still a Case of Great Expectations?

Khanna, Ashish K. MD, FCCP, FCCM, FASA*,†; Gan, Tong J. MD, MHS, FRCA, MBA

Author Information
doi: 10.1213/ANE.0000000000006001

See Article, p 900

As anesthesiologists, we take care of patients in a variety of clinical settings. At each patient touch point, we intervene with therapeutics to realize a clinical response that prevents a poor outcome. Now, more than ever, we are moving from a reactive to a predictive, proactive, and preventive intervention model of care. We can predict hypotension and the need for interventions to correct hemodynamic instability in the operating room and the intensive care unit (ICU).1,2 We can predict opioid-induced respiratory depression in the inpatient general care setting.3 The overarching goal is to maintain patients in a stable physiological milieu.

When we provide an intervention to our patients—for instance, initiate a vasopressor in anticipation of hypotension or provide an antiemetic to a patient at high risk for postoperative nausea and vomiting—we presume that the appropriate choice of therapy will translate into a desired clinical response. However, this is often not the case. Patients with vasodilatory shock with a nearly identical etiology and presentation respond differently to the same set or bundle of sepsis interventions. Some patients never experience adequate pain relief in the postoperative period despite considerable doses of opioid and nonopioid analgesics, while others do not respond to any antiemetics.

Clinical wisdom teaches us that “one size does not fit all,” and that no 2 patients (except, perhaps, identical twins) are exactly alike. Hence, we are seeking to determine how our patients are specifically different and how we can provide tailor-made, personalized therapy, which is determined by a pharmacogenetic blueprint or map. Developing and delivering therapies targeted to individuals or groups, based on their unique genetic, molecular, or phenotypic characteristics, is perioperative precision medicine. On the surface, perioperative precision medicine seems easy enough to translate to routine bedside use. However, until now, we have not realized our scientific dream of 1 drop of blood opening an array of laboratory testing modalities, which will inform and support precise therapeutics and predict optimal responses to these therapeutics.

In this month’s issue of Anesthesia & Analgesia, Nirvik and Kertai4 present a futuristic vision of perioperative precision medicine. These authors persuasively highlight the predictive, preventive, personalized, participatory, and pertinent facets of perioperative precision medicine. The National Institutes of Health (NIH) is also advocating for the precision medicine movement by building a prospective precision medicine database. However, a general reluctance to participate appears evident given that Nirvik and Kertai note that only 200,000 individuals have enrolled in this NIH database as of 2019. The role of genomics is evident, and it can be traditionally captured with genome-wide association studies (GWASs), whereby genetic variants would lead to a particular phenotype with organ system injury. However, phenome-wide association studies (PWASs) seem to be more challenging, but may be more fruitful, wherein the use of single genetic predictor could identify multiple phenotypes.

The term big data has become commonly used but loosely defined. Larger sample sizes for study cohorts are a frequently touted benefit of big data. Randomized trials are typically performed with a power analysis dictating that there is an 80% or greater probability of detecting an effect if it is truly present. This power requires larger sample sizes for relatively rare but clinically important events or outcomes. Investigators traditionally consider a ran-domized trial as the gold standard study design to base clinical guidelines and to change practice. The assumption is that these large-sample scientific experiments generate data from a study sample that presumably behaves the same way as the real-world patient population—but this is mostly untrue. Real-world clinical data can instead look very different from the best prospectively planned and most inclusive randomized trial.

Nirvik and Kertai4 herein speak to the novel “n-of-1” trial, in which the individual patient, receiving customized therapy, is the focus of the clinical research testing a study hypothesis. A chronic condition with repetitive and cyclical use of the same therapeutic intervention is a good example. A traditional trial would test a drug against a placebo once and determine the outcomes. Classifying patients into strict, binary categories of responders and nonresponders may be incorrect in this setting. The results would simply imply that at least at that instant of the trial, the experimental drug worked better or worse than placebo. If tested repeatedly, there would be variations in responses. Subsequently working on pharmacogenomic profiling to search for reasons for dichotomous outcomes is a futile exercise. A better approach would utilize repeated tests with the same drug on the same individuals to assess for subtle differences in effects each time.5

Another promising option is the alternating interventions or multiple crossover design, in which a pseudorandomized sample can be achieved at a fraction of the cost and effort of running a traditional prospective randomized trial. One such prospective alternating interventions trial is one in which patients admitted to a hospital general care unit are pragmatically enrolled into alternating 4-week blocks of continuous oxygenation and hemodynamics monitoring, which is unblinded to the provider in one cluster and blinded in the other. The blinded cluster is essentially only able to provide intermittent vital signs per traditional care. These 4-week blocks alternate and cross over during the entire year, whereby researchers can acquire real-world data and manage pseudorandomization of patients such that confounders (covariates) are nearly equally distributed (Table). Such a trial would be nearly impossible to undertake, with the once-per-patient randomized model of enrollment, which would take several years and require extensive resources.

Table. - An Example of an Alternating Interventions Cohort Trial Design
Hospitalward clusters Weeks 1–4 Weeks 5–8 Weeks 9–12 Weeks 13–17 Weeks 18–21 Weeks 22–25 Weeks 26–29 Weeks 30–33 Weeks 34–37 Weeks 38–41 Weeks 42–44 Weeks 45–48 Weeks 49–52
Ward 1 U B U B U B U B U B U B U
Ward 2 B U B U B U B U B U B U B
Abbreviations: U, unblinded to continuous monitoring and B, blinded to continuous monitoring - for the hospital ward system that will cross over every 4 wk.

Perioperative precision medicine holds promise for the future of anesthesiology and its intersection with perioperative medicine. Data suggest that if viewed as a distinct disease entity, mortality within 30 days after noncardiac surgery is the third most common cause of death in the United States, and the global postoperative mortality numbers are in the millions.6 In a large cohort of adults undergoing noncardiac surgery, nearly all deaths occurred after the procedure, and nearly half of these were associated with 3 complications: major bleeding, myocardial injury, and sepsis.7 Given that these 3 postoperative complications occur beyond the immediate recovery period, anesthesiologists need to focus on preventing harm and identifying patient outcomes that can be influenced by improved care. As defined by Nirvik and Kertai,4 the perioptome is the integration of the genome, surgery, anesthesia, postoperative recovery, and critical care.

There have been some recent advances in genetic-based prediction of analgesia, opioid-induced respiratory depression, nausea and vomiting, and vascular thrombosis.8–10 Nirvik and Kertai4 highlight other ongoing efforts at predicting acute kidney injury, myocardial injury after noncardiac surgery (MINS), malignant hyperthermia, and postoperative atrial fibrillation. Several other opportunities remain unexplored. Perioperative hypotension has been strongly associated with postoperative myocardial injury, and some success has been realized in predicting intraoperative hypotension. However, there is variation in patient response to fluids, vasopressors, and inotropes administered to correct hypotension. Investigation has focused on renin as a biomarker to predict shock outcomes and engage the renin-angiotensin system via exogenous angiotensin as the preferred vasopressor in some patients—the limitation being the lack of a point-of-care renin assay for these patients.11 Similarly, a deficiency of vasopressin and impaired sympathetic response from the adrenal medulla have been identified in shock states, but this phenomenon is almost never examined at the bedside before initiation of a vasopressor specific to a deficient state.12,13

Certainly, looking into the future state of the ICU of 2050, there is a vision to have a set of admission biomarkers for all patients in shock and to tailor vasopressor therapy with specific dose and duration profiles that precisely align with shock types and biomarker elevations. In our opinion, this precision medicine pathway would also include moving away from the mean arterial pressure of 65 mm Hg or more for all patients and would include indices such as continuous abdominal perfusion pressure and advanced hemodynamics, including microcirculatory flow to guide therapy.14

Similarly, on hospital general care units, extensive surveillance monitoring has identified predictors for postoperative opioid-induced respiratory depression.3 Almost a decade ago, important work identified tandem repeats of interleukin receptor antagonist gene segments that may contribute to differences in opioid consumption in the postoperative period.15 However, much still needs to be done to develop a precision-based set of biomarkers that would alert the clinician to genetic susceptibility to adverse effects from certain opioid types for certain patients. Furthermore, a dose of opioid that is titrated to subjective patient-driven pain scores is inherently plagued with several deficiencies. If nociception-based monitoring and opioid dosing are done properly and consistently for all inpatients recovering from surgery, in conjunction with genetic profiling for opioid susceptibility and portable biosensor-based monitoring for opioid-induced respiratory depression, an individual patient could achieve optimal analgesia with minimal adverse safety events.16

Considerable work continues in pharmacogenetic testing for drugs used in the perioperative period. Specifically, the cytochrome P450 system and the related family of CYP3A4 genetic polymorphisms are responsible for nearly 80% of all drug biotransformations.17 Here is an important opportunity to risk-stratify large population cohorts to a broad range of commonly provided medications, such as analgesics, antiemetics, and vasoactive drugs.18 Nirvik and Kertai4 also note the US Food and Drug Administration (FDA) identifying 258 drugs, 461 corresponding gene variants and mutations, with 20 anesthetic drugs, and 7.75% of the pharmacogenetic pool. This represents a significant opportunity for intervention that would translate into improved patient safety. Disturbingly, however, only 12% of drugs licensed in the last 2 decades carry pharmacogenetic biomarker safety labels.

All this said, what hinders us from applying this knowledge of perioperative precision medicine and turning a drop of blood from our patients into targeted therapy with predictable outcomes? Is it the lack of evidence, the quality of evidence, the need for more appropriate trial designs, or a combination of all of these? Or is it the very familiar cost-benefit equation that we almost never measure well in real-life practice?

We feel it is much more than just these usual culprits that prevent the translation of perioperative precision medicine to the bedside. It is probably a lack of prioritization and inertia, proper stakeholder engagement at the design stage of interventions, and robust dissemination and implementation of science techniques, which would make a difference in cultural perceptions and adoption of new technology.

An ideal perioperative precision medicine risk-profiling system should be all encompassing; validated in several diverse population types; integrated into the electronic medical record workflow; extensively trained and tested in the real-world hospital environment; patient friendly; cost-effective; and most important, reimbursed by insurance carriers in standard cost of care bundles. Appropriate trial methodology, more n-of-1 trials, less emphasis on dichotomous outcomes such as response or nonresponse, and more reliance on improvement-based continuous end points will promote correctly analyzing the data.

Clearly, perioperative precision medicine is here to stay. It will need to be balanced appropriately on the pillars of artificial intelligence, machine learning, big data informatics, and pharmacogenetic profiling. That is the only way we can move the needle to bring it to the bedside so we can provide all patients under our care the precise therapy they require.

DISCLOSURES

Name: Ashish K. Khanna, MD, FCCP, FCCM, FASA.

Contributions: This author helped conceive and write this invited editorial in its entirety.

Conflicts of Interest: A. K. Khanna consults for Medtronic, Edwards Lifesciences, Philips North America, GE Healthcare, Potrero Medical, Retia Medical, Caretaker Medical, and Trevena Pharmaceuticals. He is also funded with a Clinical and Translational Science Institute National Institutes of Health/National Center for Advancing Translational Sciences KL2 TR001421 award for a trial on continuous postoperative hemodynamic and saturation monitoring. The Department of Anesthesiology is supported by Edwards Lifesciences under a master clinical trials agreement. Khanna is a founding member of the BrainX group.

Name: Tong J. Gan, MD, MHS, FRCA, MBA.

Contributions: This author helped conceive and review this invited editorial in its entirety.

Conflicts of Interest: T. J. Gan consults for Acacia, Baudax, Edwards Lifesciences, Merck, and Medtronic.

This manuscript was handled by: Thomas R. Vetter, MD, MPH.

    REFERENCES

    1. Hatib F, Jian Z, Buddi S, et al. Machine-learning algorithm to predict hypotension based on high-fidelity arterial pressure waveform analysis. Anesthesiology. 2018;129:663–674.
    2. Rahman A, Chang Y, Dong J, et al. Early prediction of hemodynamic interventions in the intensive care unit using machine learning. Crit Care. 2021;25:388.
    3. Khanna AK, Bergese SD, Jungquist CR, et al.; PRediction of Opioid-induced respiratory Depression In patients monitored by capnoGraphY (PRODIGY) Group Collaborators. Prediction of opioid-induced respiratory depression on inpatient wards using continuous capnography and oximetry: an international prospective, observational trial. Anesth Analg. 2020;131:1012–1024.
    4. Nirvik P, Kertai M. Future of perioperative precision medicine: integration of molecular science, dynamic healthcare informatics, and implementation of predictive pathways in real-time. Anesth Analg. 2022;134:900-908
    5. Senn S. Statistical pitfalls of personalized medicine. Nature. 2018;563:619–621.
    6. Nepogodiev D, Martin J, Biccard B, Makupe A, Bhangu A; National Institute for Health Research Global Health Research Unit on Global Surgery. Global burden of postoperative death. Lancet. 2019;393:401.
    7. Spence J, LeManach Y, Chan MTV, et al.; Investigators TVEiNSPCES. Association between complications and death within 30 days after noncardiac surgery. Canadian Med Assoc J. 2019;191:E830–E837.
    8. Senagore AJ, Champagne BJ, Dosokey E, et al. Pharmacogenetics-guided analgesics in major abdominal surgery: further benefits within an enhanced recovery protocol. Am J Surg. 2017;213:467–472.
    9. Janicki PK, Schuler HG, Jarzembowski TM, Rossi M II. Prevention of postoperative nausea and vomiting with granisetron and dolasetron in relation to CYP2D6 genotype. Anesth Analg. 2006;102:1127–1133.
    10. Pirmohamed M, Burnside G, Eriksson N, et al.; EU-PACT Group. A randomized trial of genotype-guided dosing of warfarin. N Engl J Med. 2013;369:2294–2303.
    11. Khanna AK. Renin kinetics and mortality-same, same but different? Crit Care Med. 2022;50:153–157.
    12. Garrard CS, Kontoyannis DA, Piepoli M. Spectral analysis of heart rate variability in the sepsis syndrome. Clin Auton Res. 1993;3:5–13.
    13. Landry DW, Levin HR, Gallant EM, et al. Vasopressin deficiency contributes to the vasodilation of septic shock. Circulation. 1997;95:1122–1125.
    14. Asfar P, Radermacher P, Ostermann M. MAP of 65: target of the past? Intensive Care Med. 2018;44:1551–1552.
    15. Candiotti KA, Yang Z, Morris R, et al. Polymorphism in the interleukin-1 receptor antagonist gene is associated with serum interleukin-1 receptor antagonist concentrations and postoperative opioid consumption. Anesthesiology. 2011;114:1162–1168.
    16. Meijer F, Honing M, Roor T, et al. Reduced postoperative pain using nociception level-guided fentanyl dosing during sevoflurane anaesthesia: a randomised controlled trial. Br J Anaesth. 2020;125:1070–1078.
    17. Zanger UM, Schwab M. Cytochrome P450 enzymes in drug metabolism: regulation of gene expression, enzyme activities, and impact of genetic variation. Pharmacol Ther. 2013;138:103–141.
    18. 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.
    Copyright © 2022 International Anesthesia Research Society