Bayesian analysis: Free vs Commercial options
Free Bayesian solutions with their extremely limited Bayesian priors lack the fundamental characteristics required to safely modify the pharmacodynamic parameters of a drug such as vancomycin. When attempting to use Bayesian logic with inadequate data in a medical setting, the potential for harm is high. Dosing errors and poor patient outcomes are possible.
GlobalRPh Top Recommendation
PrecisePK is one of the leading commercial Bayesian dosing solutions. Its complex machine learning algorithms help automate the complex Bayesian analytical model building and this is one of the primary differences that separate commercial tools from the free versions.
Results
Bayesian Forecasting computing algorithm
- PrecisePK was one of the world’s first Bayesian dosing software programs. Established in 1980 and formerly known as TDMS, their program is trusted by well-known institutions such as UCSD, Rady Children’s Hospital, and others to produce reliable proven results for more than a million patients over 30 years.
- Precise PK offers an individual recommendation after just one serum concentration input. In a robust Bayesian program like PrecisePK, it actually does not matter when that drug level is obtained and you generally only need one level for the program to be able to hone in individual’s pharmacokinetics, thus giving a very precise dosing recommendation or AUC prediction
Population covered (Extensive Data Set)
Bayesian priors are derived from richly sampled and clinically validated studies that represent a wide range of population. These large databases significantly improve the statistically relevant predictive power.
For example, their vancomycin drug module alone was built on over 40 published rich pharmacokinetic data sets encompassing neonates, infants, pediatrics and adults with different comorbidities and special conditions such as obesity, ICU, burn, renal impairment, cardiac surgeries, etc.
They also consider additional drug factors such as concomitant medication used, etc. Data is filtered by demographic characteristics and further refined via built-in extra drug factors that allows the selection of the most suitable population PK model.
Models used
Their robust Bayesian platform houses different types of models including one compartment with first order elimination, two compartments with first order elimination, one compartment with Michaelis-Menten elimination and two compartments with Michaelis-Menten elimination. This allows them to capture various intricate drug disposition profiles of different drugs. For example, vancomycin PK is best described via a 2 compartment model with first order elimination and voriconazole is best described via a 2 compartment model with Michaelis-Menten elimination.
Continuous Learning
As Bayesian priors constantly evolve with new patient data and information, their local population features employ AI algorithms and machine learning to continuously refine and grow existing models based on the institution’s specific population, thus maximizing the target attainment.
Continuously learning the pharmacokinetic models specific to a hospital’s local population can help discover new parameters that influence dosing calculations.
Validation
In a recent paper published by Turner, R.B., et al. in the journal Pharmacotherapy, PrecisePK was shown to be the most accurate in predicting the Vancomycin AUC in critically ill patients. The graph below shows the ratio between the empirically measured AUC (using the trapezoid rule) and the AUC predicted by the software.
Therefore, a value of 1.0 would imply the highest accuracy. PrecisePK was able to predict the AUC most accurately due to the comprehensiveness of the pharmacokinetics models and the two-compartment kinetics models used by the program in its computations.
Turner RB, Kojiro K, Shephard EA, Won R, Chang E, Chan D, Elbarbry F. Review and Validation of Bayesian Dose-Optimizing Software and Equations for Calculation of the Vancomycin Area Under the Curve in Critically Ill Patients. Pharmacotherapy. 2018 Dec;38(12):1174-1183. doi: 10.1002/phar.2191.
Why Bayesian-Guided Dosing?
- Safer and more accurate vancomycin serum level predictions compared to formula-based methods, such as trapezoidal method.
When administered in excess, vancomycin can lead to acute kidney injuries (AKIs) and nephrotoxicity. Evidence shows that predicted vancomycin concentrations using Bayesian method were more accurate and closer to actual concentration levels as compared to using formula-based methods.
- Improves logistical workflow by saving time and reducing human computational effort.
The Bayesian approach to AUC estimation requires as little as one serum level to calculate AUC while manual calculation of AUC requires two serum levels. Manual calculation can also be time-consuming and computationally complex. Traditional trough-based methods of monitoring require time-sensitive blood draws to get an accurate trough level. Therefore, Bayesian method allows for more flexibility and ease in the clinical workflow.
- Highly adaptive and responsive to unique individual characteristics.
Both manual AUC calculation and trough-based monitoring only provide snapshots in time. In contrast, the Bayesian approach to dosing accounts for a patient’s physiological change throughout the entire course of their treatment. Each new serum level collected helps improve the modeling of a patient’s unique PK parameters, further improving dosing recommendations and predictions.
- Save on immediate costs such as blood draws and long-run costs such as cost of nephrotoxicity incidence, extended hospital stay and more.
Collecting multiple serum levels not only adds inconvenience to clinical pharmacists’ workflow but also potentially increases operational cost. In the long-run, there could be added costs related to inaccurate dosing regimens using the traditional methods such as nephrotoxicity and AKI treatment. These unnecessary costs could be avoided using Bayesian-guided dosing.
EMR Integration
Unlike other Free software programs, PrecisePK is fully integrated with the Epic EHR platform on the App Orchard Marketplace. The integration allows them to streamline patient care and provide the health system with direct access to the Bayesian Guided Precision Dosing program. When launching Precise PK within the Epic environment, the lifecycle information of the medication orders such as previous doses, and patients’ vancomycin levels and serum creatinine will be securely and automatically imported into PrecisePK software and incorporated into the Bayesian algorithm. Ultimately, the Epic integration will allow optimization of user experience, reduction of pharmacy workload, reductions in the transcription of error, quick and accurate dosing calculation and interpretation, and seamless coordination of care.
Collaboration Notes
Availability of documentation and note taking features promotes coordination of care and ensure accurate communication between different medical team members
All data is handled and stored according to the HIPAA guidelines. PHI collected by the software is automatically stored in an encrypted database without human involvement. Their database is highly secured, fully encrypted and automatically backed up, regardless of whether the data is in-transit or at rest
Dose Analytics and Flexible customer optimization
Development of the dose analytics dashboard ensures the therapeutic use and monitoring is safe and effective across different populations and medications.
Marketable differences between Precise PK and Doseme/ Insight RX
- Full access to a complete drug library, at no extra charge
- “Least biased and most accurate”
- A highly flexible and supportive team.
- 24/7 customer support to assist through navigating the software.
