Research design

This is a prospective, single-centre pilot study collecting health-related data and blood samples of patients suffering from pain and having an indication to switch from oral or intravenous morphine to transdermal fentanyl. Transdermal fentanyl administration is observed during 3 days (72 hours) for study purposes. Any changes in individual responses observed during transdermal fentanyl administration (including clinical signs of overdose/toxicity) are recorded and used to potentially improve DT-assisted transdermal fentanyl dosing. In addition, we will study the association between individual fentanyl PK, pain response and toxicity. The study is not expected to impact the clinical routine or clinical decision-making processes, including the initial dosage and dosage adjustments. The DT will not make fentanyl dosing recommendations for patients included in this clinical study, the clinician will, based on the Opimeter tool and clinical assessment. In parallel, we ran the DT simulation to predict the optimal initial concentration, given the patient-specific data and previous opioid dosage that provided a stable analgesic effect.

Patients participating in the study will be monitored closely (blood samples, continuous cardiorespiratory monitoring, pain feedback), which increases patient safety for those participating in the study.

This research project will be conducted in accordance with the protocol, the Palliative Care Formulary,49 the Declaration of Helsinki,50 the principles of Good Clinical Practice, the Human Research Act and the Human Research Ordinance51 as well as other locally relevant regulations.

Study participants

We wrote a Protocol and submitted a standard ethical application to the cantonal Ethics Committee of Eastern Switzerland (EKOS). For triage among palliative care Kanton Spital Sinkt Gallen (KSSG) unit patients, we formulated inclusion criteria: participants include 18 years or older, their life expectancy should be longer than 3 months and all participants should be on oral or intravenous opiate therapy (with or without co-analgesics) with an indication for a switch to transdermal fentanyl, and they should have a stable pain score (below 3 according to VAS) for at least 24 hours before switching to transdermal fentanyl. The type and dose of comedications are documented at baseline and during the study period, as they may impact pain control. Potential adverse skin reactions (categories: skin disorders: pruritus, rash, hyperhidrosis; skin and subcutaneous tissue disorders: eczema, allergic dermatitis, contact dermatitis) are all documented at baseline and during study treatment. Participants should have an intravenous catheter in place, allowing blood samples to be taken and they should have at least two more hospitalisation days planned. Major exclusion criteria include an intrathecal catheter for pain treatment, major surgery, dialysis, cognitive dysfunction, skin disease or excessive sweating affecting transdermal drug absorption, known patch allergies, intended epidural analgesia(PDA).

All patients hospitalised at the palliative or oncology ward of the study site receiving oral or intravenous opioids for at least 2 days will be screened for potential participation in this study. All eligible patients will be asked for their participation in the study after careful information on the purposes of the study. Prior to registration, the following steps have to be performed: checking the eligibility criteria, checking informed consent from the patient prior to any protocol-specific procedure and filling in the patient’s screening, enrolment and identification lists. Only authorised study investigators will inform patients and collect informed consent. For this purpose, a staff list with respective authorisation for specific study tasks is kept in the local main study files. The study investigator will first discuss with the responsible inpatient staff whether specific patients may fit for the study purpose. If positive, the study investigator is responsible for explaining the purpose and procedures of the study to a specific patient. Every patient will get at least 24 hours for considering study participation. The study investigator is also responsible to clarify potential study-related questions with each specific patient before signing the Participant Information and Consent Form (PICF). We have established a study-specific form that documents the date of first patient contact, the date of clarification of potential study-related questions and the date of signing the PICF.

No stratification is to be performed. Registration will be done at the Kantonsspital St. Gallen, and all data will be recorded according to the protocol. All involved investigators and the project leader of the study will regularly be updated on patient accrual. In addition, we defined criteria for withdrawal and discontinuation: participating patients have the right to withdraw from the study at any time without having to specify the reason for withdrawal. In case of withdrawal, the biological material and health-related personal data collected until the withdrawal date will be further used and stored in pseudo-anonymised form on the termination of data analysis. For participants who did not give consent for further use of their biological materials, all collected plasma samples will be destroyed on completion of this research project. According to a priori power calculation, this study will include 19 participants.

Pretreatment evaluations and procedures

Written informed consent must be obtained within 2 days prior to starting transdermal fentanyl and documentation of a chronic pain problem currently requiring oral or intravenous opioids.

The following investigations are mandatory within 2 days before registration: medical history including smoking history, baseline pain score and physical examination (including blood pressure, WHO performance status, weight, Body Surface Area (BSA), lean body mass) and documentation of individual medications, including the dose and interval of main analgesics and co-analgesic drugs.

Fentanyl therapy

Transdermal fentanyl is given according to the treating palliative physician or oncologist. We are referring to the newest recommendations and the official Swiss website (www.kompendium.ch) for product information and detailed information on handling and safety. Transdermal fentanyl may be given in combination with co-analgesic drugs. Co-analgesics should not be changed within 3 days before and after switching to transdermal fentanyl. Morphine hydrochloride immediate release can be taken as ‘PRN’ medication in case of registered pain, and documented accordingly. Recommended transdermal fentanyl dose is taken from University Hospital of Zurich’s Opimeter and rounded to ±12 µg of transdermal fentanyl. Transdermal fentanyl patches (50 µg and 75 µg are typically used) are renewed every 3 days unless early removal is indicated for fentanyl-associated toxicity. Transdermal fentanyl dose increase for inadequate pain control is recommended with the second patch application.

Transdermal fentanyl adverse events

An adverse event (AE) is any unfavourable and unintended sign (including an abnormal laboratory finding), symptom or disease temporally associated with the use of a medical treatment or procedure that may or may not be considered related to the medical treatment or procedure. Patients will be instructed by the investigator to report the occurrence of any AE. The investigator assesses and records of all AEs observed during the AE reporting period of 3 days. AEs are coded with the National Cancer Institute (NCI) Common Terminology Criteria for Adverse Events (CTCAE) V.5.0, and assigned a grade (from 1=mild to 5=death related to AE) as well as a relationship to trial treatment. The NCI CTCAE V.5.0 (as pdf) as well as instructions on how to use the criteria can be found in https://ctep.cancer.gov/protocoldevelopment/electronic_applications/docs/CTCAE_v5_Quick_Reference_5x7.pdf. The most commonly reported and/or clinically relevant treatment-related adverse drug reactions for transdermal fentanyl are gastrointestinal disorders32 (especially obstipation, nausea, vomiting), central toxicity (vertigo, cephalea, fatigue, depression, anxiety, agitation), tachycardia and palpitations, respiratory depression and bradypnoea. In this research project, only AEs considered at least possibly related to fentanyl treatment are recorded.

Evaluations during treatment

Visits will take place for the whole study period of 3 days, starting from switch to transdermal fentanyl treatment. Our study visits include: physical examination (including blood pressure, WHO performance status, weight, Numerical Rating Scale and VAS pain score); AEs; documentation of individual dose and interval of transdermal fentanyl and co-analgesic drugs; and fentanyl PK assessments at 0, 8, 16, 24 and 34 (and 72) hours after the start of study treatment.

Therapeutic drug monitoring

During this observational clinical study staff will collect the blood samples of the patients who are undergoing transdermal fentanyl therapy at KSSG hospital palliative care department. Venous blood sampling for fentanyl PK assessments will be done at baseline before starting transdermal fentanyl, at 8, 16, 24 and 34 (and 72) hours after the start of transdermal fentanyl using 7.5 mL serum tubes (without gel) (see separate fentanyl PK list). After sampling, blood will be centrifuged (2147 G, 10 min) and serum extracted at the Laboratory Kantonsspital St. Gallen. The tubes will be stored at −80°C until the end of the study. At the end of the study, blood samples will be sent in batches on dried ice to the Center for Laboratory Medicine of Inselspital. The method that is going to be used for the determination of fentanyl in blood samples/plasma collected within this study is a simple liquid chromatography-tandem mass spectrometry which determines plasma concentrations by deproteinisation of sample with acetonitrile. Notably, the required plasma volume for detection in this method is reduced in comparison to standard methods, to just 20 (µL). The previous study52 demonstrated good calibration curves in concentration range from 0.05 to 5 ng/mL. Sample processing will be performed using an automated, routinely used pipeline on the Tecan Evo-150. The analysis will be performed using a validated and accredited in-house method on the Shimadzu LCMS-8060nx. The latter has a lower limit of quantification for fentanyl of 0.15 µg/L.

Observations

Comparison of predicted dosage for transdermal fentanyl application based on Opimeter and DT (input parameters: age, weight, gender, height, glomerular filtration rate, actual opioid dose, aspartate aminotransferase-AST, bilirubin, creatinine, albumin and C-reactive protein-CRP) (single observation as baseline). Comparison of the observed peak plasma fentanyl concentration and its time point with those predicted by the digital twin for the applied dose in the clinic (the required input parameters are listed in the simulations section). However, it is important to acknowledge that due to the limited number of data points collected, there is a possibility that some peaks may be overlooked. The reason is that between two data points, an additional peak in concentration can occur, which would be picked up if the temporal resolution would be higher.

In addition, we would perform the comparison of intermittently recorded patient data according to the clinical routine with continuous textile-sensor-based readings to detect drug-induced changes in respiratory responses.

Sample size

An acceptable tolerability margin of ±20% (for intraindividual variability) for comparing individual fentanyl C_max plasma concentrations, as calculated by the PK-model, with DT-predicted values results in a bivariate correlation coefficient of r=0.97 (H1). In contrast, a tolerability margin of ±30% (r=0.91) is not considered acceptable (H0). A power and sample size analysis for z tests (correlations: two dependent Pearson r’s; G*Power V.3.1.9.7) revealed a required sample size of n=19 to achieve a statistical power (1-β error) of 80%. Besides known high interindividual variability, variation can be caused by errors in the mathematical model, wrong inputs for the model, problems in our assumptions and oversimplifying of the model.

Data management and analysis

At the beginning of data collection (after the clinical triage of inpatients for the study) informed consent communication and signing were done.

The next step is data acquisition from inpatients (age, weight, gender, height, glomerular filtration rate, actual opioid dose, pain level VAS, AST, bilirubin, creatinine, albumin and CRP), followed by the calculation of initial fentanyl dosage based on Opimeter calculation. Patients are then receiving the therapy based on a decision that the clinician made, taking into account Opimeter calculation and her knowledge and experience, according to the clinical routine. The clinical part of the team will then communicate these data with the simulation team in a safe and standardised way (GDPR—The General Data Protection Regulation—compliant) where each participant is filed under a unique number in the study, his/her attending physician being the only person with the possession of the identity information.

The attending physician inputs patient’s data into the Empa mobile application (ie, linked to the previously developed fentanyl DT model) to get the plasma concentrations, pain levels, minute-breathing and clearance rates calculated by the DT model. In the end for every included participant the team will have three data points: the Opimeter calculation, the levels of fentanyl plasma concentrations and DT calculations. Our primary aim is to confirm how good DT prediction versus actual plasma concentrations was, at the end of the study.

• First we compare the accuracy of the DT predicting initial fentanyl dosage.

• In a second step, we validate the predictive performance of the DT for predicting the fentanyl plasma concentration (for this validation, we use the same initial fentanyl dosage as it was administered to the patient).

The output of these comparisons will be how well DT performed in contrast to real values obtained during the standard protocol of transdermal fentanyl therapy. With this comparison, Empa team members will work on further optimisation and do the post-processing of the results in order to formulate the overall improvement of the process (calibration of DT model parameters).

Patient and public involvement

None.

Simulations

Transdermal Fentanyl Delivery Digital Twin was developed earlier by our team, by implementing a physics-based model for drug penetration into the skin, a PKs model to calculate the fentanyl concentration in plasma and a PDs model for calculating the therapeutic drug effects.43–46 The DT is a virtual representation of the TDDS and the patient, using in silico calculation of the physical processes occurring during drug release, penetration, percutaneous absorption, distribution, metabolisation and excretion of the drug and clinical response, including pain relief (PDs). This DT was based on the virtual patient physiological features to explore the impact of patient’s feedback in the outcome of the therapy. In our previous work, we developed and validated mechanistic skin model43 44 for fentanyl diffusion, storage and partitioning in the epidermis to quantify drug release from the patch and percutaneous drug absorption.44 45 Then the mechanistic skin model was coupled with a PKs and a PDs model for developing a DT predicting patient’s responses (pain perception, respiratory suppression) to transdermal fentanyl application.45 It was demonstrated in silico that we could effectively tailor the duration of fentanyl patch application for virtual patients in different age categories to achieve more effective pain relief while avoiding fentanyl overexposure.44 Consequently, we showed that the DT-guided transdermal fentanyl administration was more efficient and safer when including virtual patient’s feedback on pain.45 In the next phase, we developed a virtual population (n=3000) by implementing Markov chain Monte Carlo over a modest sample size.46 We tailored our physics-based DT to the characteristics of individual patients of the virtual population to study interindividual variability in clinical outcomes. In addition, in recent theoretical work on mass-transport that is crucial for TDDS, we showed that skin’s fractal properties should be included in the mathematical description of transport processes (of drug molecules) and this highlighted important physiological implications.7 Our theoretical work on the model of the physical process in which the molecules are penetrating skin layers47 suggesting that the classical Fickean approach for inhomogenous physiological tissue should be corrected.48

The current digital twin initially evaluates the morphine and its metabolite concentration in plasma by a PKs model for oral morphine. In the next step, by the PDs model, the morphine effects, both pain relief and reduction in minute ventilation, is calculated. Fick’s second law was implemented to calculate the penetration of fentanyl throughout the skin to reach blood circulation. Then, using the PKs model, the fentanyl concentration in plasma is calculated and based on this concentration, the PDs model calculates the pain relief and reduction in minute ventilation. The digital twin was developed in COMSOL Multiphysics (V.5.6, COMSOL AB, Stockholm Sweden). COMSOL is commercial finite-element-based software. During the building of the model, the model coefficients were based on experimental results from the literature,53 54 as well as updated in line with the computational framework.55 The model and simulation were built and executed according to the best practice guidelines in modelling for medical device design.56 57 Furthermore, the digital twin was tailored to the physiological features of the virtual patients. The tailoring was done by modification of model parameters corresponding to the age, weight, gender, height, serum bilirubin, serum creatinine and serum albumin of the virtual patients. In this way, the digital twin was able to predict the outcome of opioid therapy uniquely for each virtual patient.

The main outcomes of the model are: C__max and t__max (maximum concentration of fentanyl in plasma and time to reach the maximum concentration), time lags (for plasma concentration and for effect), pain intensity, minute ventilation, concentration of fentanyl in each organ, concentration of metabolites and the amount of cleared drug from the body.

As in every model built in COMSOL, it is possible to export the mobile application that can be used as non-expert in modelling for practical reasons. In case of this DT, an appropriate mobile application was built (see below).

We want to stress here that no patient would get DT suggested initial dosage, it would be just a numerical value resulting from running the model equations that were proven in pharmacological science to model adequately physiological processes. The clinician will always make the decision, as the digital twin is not yet validated nor approved, validation is the goal of this pilot. Based on the knowledge extracted from this experiment, the model would be improved, that is, parameters would be calibrated, and the overall process would be optimised to accommodate the differences measured here (inverse engineering).

Statistical analysis

Whenever we use simulation (as described above), its output is given as a graphic representation and calculated parameter values are given in tables. These numerical data can be used for post-processing (exported as an Excel table and used in another statistics software), in this case, statistical analysis. What we want to explore is how big the difference is between measured values (eg, fentanyl plasma concentrations) obtained from the experiment (biochemical analyses of blood samples) for each patient in already specified time intervals during the 3 days of the TDD therapy and those calculated by the model, in order to further improve the simulations. For any statistical analysis, we intend to use R and RStudio which is free software used in our prior research.

Dedicated mobile application for clinicians (a possible extension)

A mobile application is developed that is offering to clinicians a graphical representation (together with a table with all the numerical values), and with simple installation of its exe file one can start using it like any other application, locally. This application is at this point a mockup to be able to explore this with clinicians and is not implemented in the clinical workflow. Standardised clinical protocol (for TDD fentanyl) applies, and all the comparisons are done at the end of the study to confirm the differences between DT calculations and reality. From a technical point of view, both the patient and clinician do not even have to know what DT is calculating, as this is not intended to inform the current therapeutic process, only to provide data for post-process validation. To summarise, DT is not participating decision process, it is still a standard clinicians’s decision.

An illustration of what is possible to get with the mobile application once a clinician enters the data about the actual patient (previous morphine dose that kept constant analgesia level, age, sex, weight, height, pain level VAS, bilirubin, albumin and creatinine values) it is clear that s/he can have additional information about the possible biological responses. From the table below the graphs that are appearing on the screen, s/he can see whether there are some discrepancies. In this way, the clinician can also get an appreciation of the fentanyl concentration evolution and predicted pain evolution for different therapies. Such data is time-consuming to obtain experimentally, especially at a high temporal resolution.

In our previous work, we simulated those outcomes on virtual patients (the large number of individual avatars generated from a dataset of real values, that are mimicking the epidemiological data in the population) given the different initial therapy. Physics-based simulation is actually giving another insight to a clinician, prompting if the suggested Opimeter value is too far from what DT is predicting for the specific patient. This would be a rare opportunity to compare the actual fentanyl concentrations in a person’s plasma (real experimental data), Opimeter calculations and DT calculations to see the discrepancies that can lead to optimisation of the model. This first value refers to a sentence from our protocol: ‘Venous blood sampling for fentanyl PK assessments are done at baseline before starting transdermal fentanyl (at 0), at 8, 16, 24 and 34 (+72) hours after the start of transdermal fentanyl’. This means (in terms of fentanyl plasma concertation) five samples per patient during the first patch application, representing the drug accumulation (and circulation) in the organism. We are effectively constructing the curves that illustrate the time evolution of concentration and can thus compare the difference between the simulated curve and the one based on experimentally obtained values.

What will be compared here (onsite comparison by the doctor and post-evaluation by the scientist) are the levels of perceived pain (input from the patient), the concentrations of fentanyl in plasma (to be measured during the ongoing therapy) and the minute ventilation (see next section Application mockup for examples).

It is important to note that the TDD fentanyl therapy (application of transdermal fentanyl patches) is going to last for three consecutive days, and DT simulation can be performed (with adequate input about the patient in the form of his/her physiological characteristics) in a couple of minutes (up to 3 min to be precise). Of course, a simulation cannot exactly mimic the real longer lasting event, but it can give an insight important to the clinician to realise whether the initial dosage suggested by Opimeter is under or over-shoot.

Application mockup

This Application is exported from COMSOL Multiphysics in which the Digital Twin for transdermal fentanyl therapy model was developed to a standalone application (figure 1). The screenshots are illustrating what the clinician can see (curves giving information about the time evolution of certain values, plus tabular values that could be changed to various interintervals; one can opt to get from minute, to hours expected values in the table and at a curve).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Overview of the application synchronised with Digital Twin model, for clinical staff to enter relevant patient’s data, from age, weight and height, to previous therapy that secured pain management and several plasma concentrations.

Not only that a clinician can use this rapid tool to get additional information on predicted uptake and decrease of concentrations before s/he decides on the actual start of the fentanyl therapy, but the overlap between already administered morphine and fentanyl could be checked for this particular set of physiological features characterising a particular patient.

First figure is illustrating the appearance of the application exported from Digital Twin model; second figure (figure 2) shows the graph of fentanyl plasma concentrations changes over time; third (figure 3) is the temporal evolution of pain levels (as a consequence of the therapy) for this patient, and fourth figure is illustrating minute-ventilation values over time. Below each graph, the clinician would have the tabular view of numerical values used for that graph (the time step could be adjusted from minutes to hours).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Graphical representation of fentanyl concertation in plasma as a result of digital twin simulation on a set of physiological features of an individual patient. Fentanyl maximum plasma concentrations is reached at 40th hour.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

DT simulation output of the expected pain level (upper panel, A) for a particular patient, as it appears in application developed for clinicians. On the lower panel (B) minute ventilation time evolution as DT simulation output (COMSOL model application). DT, digital twin.

Our application that is connected to DT model is openly accessible.

How to use this application

The first thing a clinician can do is to fill in the patient’s data on a left-hand side of the screen (weight, age, gender, height, etc). In case of an error, on the left upper corner of the screen, there is icon ‘clear all’, or you can simply use the delete button from the tastatur to correct it. Once the information is entered, on the top left side one should click the icon ‘Run’. With that, the calculations are started. After a couple of minutes (usually below 3 min in each case), the curves could be visible on the screen. Below the curves (four different windows can be opened for further close inspection: Pain intensity, minute ventilation, morphine concentration in plasma and fentanyl concertation in plasma). In the first column (of the tables below the curves, the detailed numerical values can be seen) it is possible to change the intervals in which those values are represented. Maximal concentration C_max is simply the highest value on the curve for plasma concentration.