Patients in need of individualized precision pharmacy service often use drugs with narrow therapeutic windows, individual differences in pharmacokinetics and/or nonlinear pharmacokinetics, risk of serious adverse reactions, and genetic differences that may affect the efficacy or safety of drugs (Fig. 1). In addition, polypharmacy increases the risk of drug interactions, and the pharmacokinetics of drugs often change significantly in critically ill patients, pregnant women, children, and patients with hepatic and renal insufficiency.4–7 Patients in need of precision pharmacy service mainly include:

Fig. 1.

Flow chart of individualized precision pharmacy services. SOP: Standard operating procedures; TDM: Therapeutic drug monitoring; PGx: Pharmacogenomics; ppk: Population pharmacokinetics.

(0,35MB).

a. Patients with organ or bone marrow/stem cell transplantation: They need long-term use of immunosuppressants with narrow therapeutic window and obvious adverse reactions, and most immunosuppressants are not only substrates of liver drug enzymes and/or transporters, but also inhibit liver drug enzymes and/or transporters, which are prone to drug interactions.8

b. Patients with severe infection, including ICU patients, patients with sepsis, and patients infected with multi-drug resistant bacteria. With the rapid change of the patients' PK/PD data and their condition, the dosage and frequency of medication should be determined according to the PK/PD data.9

c. Patients taking psychotropic drugs for a long time, which can cause multiple adverse reactions and interactions. The efficacy of such drugs varies greatly among individuals with uncertain adherence.10

d. Patients taking antineoplastic drugs, which can cause severe adverse reactions with high treatment costs, and may lead to drug interactions.11

e. Patients taking multiple drugs may have drug interactions, which significantly affects the efficacy of drugs and increase adverse reactions.12

f. Others including elderly patients, children, pregnant women, and patients with abnormal liver and kidney function who may have changes in pharmacokinetics and are prone to adverse reactions.

TDM is based on individualized drug treatment with a customized drug administration plan for the patient made by measuring the patient's exposure to drugs, and pharmacological markers or pharmacodynamic indicators, applying the models of quantitative pharmacology and population pharmacokinetics, and using the drug therapeutic window as the baseline.15,16

The basic conditions for precision pharmacy service through TDM include qualified testing technology, professional pharmacists, patients who meet the monitoring indications, and customized and precise drug treatment plans.17

The analytical techniques to determine the drug concentration in biological samples (blood drug level, urine drug level, other tissue fluid or homogenate drug level) mainly include spectral analysis, chromatographic analysis, liquid chromatography-mass spectrometry, and immunological detection technology. Liquid chromatography-mass spectrometry (LC–MS) and high performance liquid chromatography (HPLC) are recommended according to the specificity of drugs. Chromatography-mass spectrometry is used more and more to conduct TDM in Chinese medical institutions.

Currently, the drugs for TDM mainly include cyclosporine A, tacrolimus, sirolimus, mycophenolic acid, carbamazepine, phenytoin, valproic acid, levetiracetam, methotrexate, vancomycin, teicoplanin, voriconazole and its metabolites, itraconazole and its metabolites, posaconazole, amikacin, meropenem, imipenem, tigecycline, linezolid, Imatinib and its metabolites, dasatinib, and nilotinib.

In addition to perform therapeutic drug monitoring of small molecule agents, we are trying to monitor macromolecule targeted agents and biosimilar agents in recent years.18

PGx is to explore the influence of genetic variation on the safety and efficacy of drugs from the perspective of genome. PGx detection helps to better explain individual differences in drug efficacy and adverse reactions,19 which brings the following advantages in terms of drug therapy:

a. Select drugs before treatment to avoid ineffective treatments, saving time for patient.

b. Predict the initial dose for each patient to help quickly determine the drug efficacy and ease the symptoms.

c. Avoid serious adverse drug reactions, such as the detection of genes related to severe skin adverse reactions of allopurinol and carbamazepine to reduce such reactions and improve medication safety of patients.20

d. Provide a scientific basis for the development and post-marketing evaluation of new drugs in clinical trials, which based on PGx.21

More commonly used methods for detecting drug-related gene mutations include fluorescence quantitative PCR, in situ hybridization, gene chip and direct sequencing, and some popular second-and third-generation sequencing methods are also used in genetic testing of drugs, especially tumor-targeted therapy.22

The drugs involved in PGx testing mainly include warfarin, clopidogrel, aspirin, methotrexate, allopurinol, tacrolimus, glucocorticoids, azathioprine, paclitaxel, antidepressants, voriconazole, proton pump inhibitors, tamoxifen, cyclophosphamide, and folic acid.23

In current medical research, real-world data from clinic is an important source of big data. Compared with clinical trials, real-world research is closer to clinical practice. TDM and PGx data are used in the analysis of patient-related information through HIS system, and an individualized medication model can be built with artificial intelligence (AI) algorithms such as machine learning. For example, the vancomycin dose prediction model has been built with the support of XGBoost machine learning algorithm to solve the problem that the dose demand varies greatly among individuals and to optimize the vancomycin dosing schedule.24 The prediction model of warfarin maintenance dose based on LightGBM machine learning algorithm can help solve the problem of missing values in the data set, and recommend the accurate dosing for patients in the case of a narrow warfarin treatment window, thus increasing the success rate of treatment.25 In addition, deep learning algorithms are increasingly used to process high-dimensional, complex and interactive clinical data. For example, the use of TabNet deep learning algorithm helps to establish the dose prediction model of lapatinib, and dig deeply into the relevant variables that affect the drug dose, establish a concise and accurate individualized medication model, and recommend the accurate dosage in clinic.26 Establishing an individualized medication model of tacrolimus with TabNet deep learning algorithm can effectively solve the problems of narrow therapeutic window of tacrolimus and individual differences, and improve the prediction accuracy of tacrolimus dosage in kidney transplant patients.27 Dose prediction models relying on machine learning and deep learning technology demonstrate good prediction performance and robustness. With the increase of sample size of input data, the prediction model will be continuously optimized to achieve better performance and practicability to support clinical decision-making and precision pharmacy service.

More elaborate pharmacy service can be provided through post-marketing observation and real-world studies. For example, the post-marketing observation clinical research of enalapril plus folic acid, showed that the progress rate of chronic kidney disease (CKD) and the decline rate of estimated glomerular filtration rate (GFR) of patients treated with enalapril plus folic acid are lower than those of patients treated with enalapril alone. Enalapril plus folic acid can reduce the risk of chronic kidney disease progression in CKD patients by 56%, the risk of eGFR decline rate by 44%, and the serum homocysteine level by 21% compared with baseline. The study showed that enalapril plus folic acid have a greater protective effect on kidney and has higher safety and lower cost. At the same time, this study also provided a basis for the management of refined pharmacy service of CKD patients with serum homocysteinemia.28

Patients with chronic diseases, especially elderly patients, often suffer from several diseases and need multiple medication treatments, which can lead to drug interactions and adverse events. Therefore, a risk assessment and early warning model for polypharmacy in the elderly is under construction through information technology to help identify such risks and provide precision pharmacy services to elderly patients.

In addition to the prescription drugs, the evaluation and management of polypharmacy should also consider the over-the-counter drugs, traditional medical drugs, herbs and health care products taken by patients at the same time.29

High-alert drugs refer to drugs that can lead to serious adverse reactions when used improperly. In order to improve medication safety, the monitoring and management of high-alert drugs through information systems are being explored in the hospital, including the management of blood sugar, pain, anticoagulants, antimicrobials and anti-tumor drugs. The high-risk groups identified by the information system are provided with precision pharmacy service. For example, for targeted anti-tumor preparations, the suitability of targeted drugs is screened through pharmacogenetic testing technology with the drug dosage for each patient clarified and possible adverse reactions predicted and monitored. For drugs with therapeutic window, the TDM technology can be used to make fine dose adjustment during the medication process, and help identify the risk of adverse reactions and adherence.