Clinical Methodology & Evidence Base

Comprehensive documentation of our Bayesian pharmacokinetic modeling platform for precision antimicrobial dosing

Table of Contents

1. Platform Overview

Purpose & Scope

This Bayesian pharmacokinetic (PK) drug dosing platform provides evidence-based, individualized dosing recommendations for antimicrobial agents requiring therapeutic drug monitoring. The system combines population pharmacokinetic models with patient-specific data to optimize dosing regimens while minimizing toxicity risk.

Clinical Applications

  • Vancomycin: AUC-guided dosing (target 400-600 mg·hr/L) for improved efficacy and reduced nephrotoxicity
  • Aminoglycosides: Extended-interval and traditional dosing strategies with peak/trough monitoring
  • Therapeutic Drug Monitoring: Bayesian estimation of individual PK parameters from sparse sampling
  • Renal Dysfunction: Dynamic creatinine clearance algorithms for patients with changing renal function

Target Users

  • Clinical pharmacists specializing in infectious diseases
  • Antimicrobial stewardship program teams
  • Critical care and infectious disease physicians
  • Pharmacy residents and fellows

2. Clinical Methodology

Bayesian Pharmacokinetic Approach

Our platform implements Bayesian Maximum A Posteriori (MAP) estimation, which combines population PK parameters (prior knowledge) with individual patient measurements (observed data) to generate personalized parameter estimates.

Bayesian Objective Function

Total Objective = Data SSE + Prior Penalty

Data SSE: Σ[(Cobs - Cpred)² / (CV × Cobs)²]

Prior Penalty: Σ[(θindividual - θpopulation) / (CV × θpopulation)]²

Where CV = coefficient of variation (typically 10% for measurements, 30% for parameters)

Core Pharmacokinetic Equations

One-Compartment Model (Vancomycin, Aminoglycosides)

Elimination Rate Constant:

ke = CL / Vd

Concentration at Time t (IV Infusion):

C(t) = (Dose / Vd) × [(1 - e-ke·tinf) / (ke × tinf)] × e-ke·(t - tinf)

where tinf = infusion duration

Steady-State Concentration:

Css(t) = (Dose / Vd) × [(1 - e-ke·tinf) / (ke × tinf)] × [e-ke·(t - tinf) / (1 - e-ke·τ)]

where τ = dosing interval

Area Under the Curve (24 hours):

AUC24 = (24 / τ) × (Dose / CL)

Half-Life:

t½ = 0.693 / ke

Two-Compartment Model (Vancomycin - Goti/Carreno)

Hybrid Constants:

α = 0.5 × [(k10 + k12 + k21) + √((k10 + k12 + k21)² - 4·k10·k21)]

β = 0.5 × [(k10 + k12 + k21) - √((k10 + k12 + k21)² - 4·k10·k21)]

Micro-Constants:

k10 = CL / Vc

k12 = Q / Vc

k21 = Q / Vp

Concentration at Time t:

C(t) = A × e-α·t + B × e-β·t

where A and B are coefficients dependent on dose, infusion rate, and micro-constants

Structural Parameters:

  • Vc = Central volume of distribution
  • Vp = Peripheral volume of distribution
  • Q = Inter-compartmental clearance
  • CL = Total body clearance

Time-Varying Renal Function Algorithms

Our platform implements three algorithms for handling changing creatinine clearance:

  1. Step-wise (default): CrCl changes instantaneously at each SCr measurement
  2. Dynamic: Piecewise linear SCr interpolation between measurements
  3. Unstable MS Chow: Secant method using two most recent SCr values for rapidly changing renal function

3. Supported Drugs & Models

Vancomycin

Model Compartments Population Target Monitoring
Matzke-Pai (MP) One General adult inpatients AUC24 400-600 mg·hr/L Peak (2hr post-infusion) + Trough
Goti 2018 Two Non-dialysis adults AUC24 400-600 mg·hr/L Peak + Trough
Carreno 2017 Two Obese adults (BMI ≥40) AUC24 400-600 mg·hr/L Peak + Trough

Aminoglycosides (Gentamicin, Tobramycin, Amikacin)

Drug Model Strategy Target Peak Target Trough
Gentamicin Matzke-Pai (One-compartment) Traditional: 1.5 mg/kg q8-24h
Pulse: 5 mg/kg q24-48h		 17.5 mg/L ≤0.3 mg/L (pulse)
0.5-2 mg/L (traditional)
Tobramycin Matzke-Pai (One-compartment) Traditional: 1.5 mg/kg q8-24h
Pulse: 5 mg/kg q24-48h		 17.5 mg/L ≤0.3 mg/L (pulse)
0.5-2 mg/L (traditional)
Amikacin Matzke-Pai (One-compartment) Traditional: 5 mg/kg q8-24h
Pulse: 15 mg/kg q24-48h		 60 mg/L ≤0.9 mg/L (pulse)
5-10 mg/L (traditional)

4. Population Pharmacokinetic Models

Matzke-Pai Model (Vancomycin & Aminoglycosides)

Volume of Distribution:

Vancomycin: Vd = 0.65 L/kg × Dosing Weight

Aminoglycosides: Vd = 0.25 L/kg × Dosing Weight

Elimination Rate (Vancomycin):

ke = 0.00107 × (CrClnorm) + 0.005216

where CrClnorm = CrCl × 1.73 / BSA (mL/min/1.73m²)

Elimination Rate (Aminoglycosides):

ke = 0.0026 × CrCl + 0.014

where CrCl in mL/min (individualized, not normalized)

Clearance:

CL = ke × Vd

Goti Two-Compartment Model (Vancomycin)

Reference: Goti V, et al. Therapeutic Drug Monitoring 2018;40(1):83-91

PubMed ID: 29095809

Population Parameters:

  • Vc = 0.8343 L/kg × Dosing Weight (central volume)
  • Vp = 38.4 L (peripheral volume)
  • Q = 6.5 L/hr (inter-compartmental clearance)
  • CL = 0.0375 L/hr per mL/min CrCl (slope), 0 L/hr (intercept)

Study Population: Non-dialysis adult inpatients receiving vancomycin

Validation: 143 patients, 397 vancomycin concentrations

Carreno Two-Compartment Model (Vancomycin - Obesity)

Reference: Carreno JJ, et al. Antimicrob Agents Chemother 2017;61(4):e02453-16

PubMed ID: 28115349

Population Parameters (Model 4 - BMI ≥40):

  • Vc = 0.234 L/kg × Dosing Weight (scaled from 25.76 L at 110 kg reference)
  • Vp = 40.96 L (peripheral volume)
  • Q = 58.99 L/hr (inter-compartmental clearance)
  • CL = 0.036 L/hr per mL/min CrCl (slope), 0.18 L/hr (intercept)

Study Population: Morbidly obese adults (BMI ≥40 kg/m², mean 110 kg)

Validation: 12 patients, 56 vancomycin concentrations

Physiologic Calculations

Body Surface Area (Mosteller Formula):

BSA (m²) = √[(Heightcm × Weightkg) / 3600]

Lean Body Weight (Devine Formula):

Males: LBW = 50 kg + 2.3 kg per inch over 5 feet

Females: LBW = 45.5 kg + 2.3 kg per inch over 5 feet

Dosing Weight:

If IBW < TBW ≤ 1.3×IBW: Dosing Weight = TBW

If TBW > 1.3×IBW: Dosing Weight = IBW + 0.4×(TBW - IBW)

Creatinine Clearance (Cockcroft-Gault):

Males: CrCl = [(140 - Age) × LBW] / (72 × SCr)

Females: CrCl = 0.85 × [(140 - Age) × LBW] / (72 × SCr)

where SCr in mg/dL, LBW in kg, CrCl in mL/min

5. Clinical Variables & Covariates

Required Patient Demographics

Variable Units PK Impact Clinical Significance
Age Years Creatinine clearance calculation (Cockcroft-Gault) Renal function declines with age; affects elimination
Sex Male/Female 0.85 multiplier for females in CrCl calculation Physiologic difference in muscle mass affects SCr baseline
Weight Kilograms Volume of distribution (Vd ∝ weight) Larger patients have greater distribution space
Height Inches or cm BSA calculation (CrCl normalization), LBW calculation Body size affects drug distribution and renal function

Laboratory Values

Measurement Units Frequency Purpose
Serum Creatinine (SCr) mg/dL Daily (recommended) Real-time renal function assessment; dynamic CrCl calculation
Drug Concentrations mg/L (µg/mL) Per protocol Bayesian parameter estimation; dose optimization

Clinical Conditions Tracked

  • Renal dysfunction: CKD stages (G1-G5 based on normalized CrCl)
  • Obesity: BMI categories (underweight, normal, overweight, obese I/II/III)
  • Amputations: Above/below knee, hip disarticulation (affects Vd)
  • Edema states: Anasarca, ascites (may increase Vd)
  • Critical illness: Burns, septic shock, febrile neutropenia (altered PK)
  • Spinal cord injury: Paraplegia, quadriplegia (affects LBW calculation)

Dosing History Requirements

  • Dose amount (mg)
  • Infusion duration (hours)
  • Dosing interval (τ, hours)
  • Date/time of administration
  • Route (IV only for included drugs)

6. Evidence Base & References

Primary Literature

Vancomycin
  1. Goti V, Chaturvedula A, Fossler MJ, Mok S, Jacob JT. Hospitalized Patients With and Without Hemodialysis Have Markedly Different Vancomycin Pharmacokinetics: A Population Pharmacokinetic Model-Based Analysis. Ther Drug Monit. 2018;40(2):212-221. PubMed: 29095809 DOI: 10.1097/FTD.0000000000000490
  2. Carreno JJ, Lomaestro B, Tietlens M, Lodise TP. Pilot Study of a Bayesian Approach to Estimate Vancomycin Exposure in Obese Patients with Limited Pharmacokinetic Sampling. Antimicrob Agents Chemother. 2017;61(4):e02478-16. PubMed: 28115349 DOI: 10.1128/AAC.02478-16
  3. Rybak MJ, Le J, Lodise TP, et al. Therapeutic monitoring of vancomycin for serious methicillin-resistant Staphylococcus aureus infections: A revised consensus guideline and review by the American Society of Health-System Pharmacists, the Infectious Diseases Society of America, the Pediatric Infectious Diseases Society, and the Society of Infectious Diseases Pharmacists. Am J Health Syst Pharm. 2020;77(11):835-864. PubMed: 32191793 DOI: 10.1093/ajhp/zxaa036
Aminoglycosides
  1. Nicolau DP, Freeman CD, Belliveau PP, Nightingale CH, Ross JW, Quintiliani R. Experience with a once-daily aminoglycoside program administered to 2,184 adult patients. Antimicrob Agents Chemother. 1995;39(3):650-655. PubMed: 7793867 DOI: 10.1128/AAC.39.3.650
  2. Barclay ML, Kirkpatrick CM, Begg EJ. Once daily aminoglycoside therapy. Is it less toxic than multiple daily doses and how should it be monitored? Clin Pharmacokinet. 1999;36(2):89-98. PubMed: 10092956 DOI: 10.2165/00003088-199936020-00001
Bayesian Methods
  1. Sheiner LB, Beal S, Rosenberg B, Marathe VV. Forecasting individual pharmacokinetics. Clin Pharmacol Ther. 1979;26(3):294-305. PubMed: 466923 DOI: 10.1002/cpt1979263294
  2. Jelliffe RW, Schumitzky A, Van Guilder M, et al. Individualizing drug dosage regimens: roles of population pharmacokinetic and dynamic models, Bayesian fitting, and adaptive control. Ther Drug Monit. 1993;15(5):380-393. PubMed: 8249044 DOI: 10.1097/00007691-199310000-00005

Clinical Practice Guidelines

  • American Society of Health-System Pharmacists (ASHP) - Vancomycin Therapeutic Monitoring Guidelines (2020)
  • Infectious Diseases Society of America (IDSA) - Vancomycin Dosing Consensus (2020)
  • Society of Infectious Diseases Pharmacists (SIDP) - AUC-Guided Vancomycin Dosing Position Paper
  • American Thoracic Society (ATS) / IDSA - Community-Acquired Pneumonia Guidelines (Aminoglycoside Dosing)

Additional Resources

7. Clinical Guidelines Compliance

2020 ASHP/IDSA/PIDS/SIDP Vancomycin Consensus Guidelines

Platform Alignment
  • AUC/MIC Targeting: Our platform calculates and targets AUC24 400-600 mg·hr/L for serious MRSA infections
  • Bayesian Methods: Implements MAP Bayesian estimation for individualized parameter estimation from sparse sampling
  • Peak and Trough Sampling: Recommends two-level sampling (peak at 2 hours post-infusion, trough pre-dose) for accurate AUC estimation
  • Renal Function Monitoring: Dynamic CrCl algorithms adapt to changing renal function
  • Nephrotoxicity Reduction: AUC-guided dosing shown to reduce nephrotoxicity vs trough-only monitoring

Extended-Interval Aminoglycoside Dosing

Best Practices
  • Once-Daily Dosing: Platform supports pulse dosing (5 mg/kg gentamicin/tobramycin, 15 mg/kg amikacin) for Q24-48H intervals
  • Hartford Nomogram Compatibility: Algorithms calculate appropriate dosing intervals based on patient-specific elimination rates
  • Trough Monitoring: Automated validation that trough ≤0.3 mg/L (gent/tobra) or ≤0.9 mg/L (amikacin) for pulse dosing safety
  • Reduced Toxicity: Extended-interval dosing associated with lower nephrotoxicity and ototoxicity rates

Antimicrobial Stewardship Integration

  • Target Attainment Metrics: Dashboard tracks percentage of patients achieving therapeutic AUC targets
  • Dose Optimization: Identifies patients requiring dose adjustments based on PK parameters
  • Toxicity Monitoring: Flags patients with supra-therapeutic exposures at risk for nephrotoxicity
  • Quality Improvement: Population analytics enable institution-specific parameter optimization

8. Development Team

Team Roles & Expertise

Clinical Lead and Software Engineer

Marshall Pierce

Credentials: BS Pharmacy, PharmD

Background:

  • Staff Pharmacist and Clinical Pharmacist at St Mary’s Hospital Richmond VA, Clinical Coordinator at Lewis Gayle Hospital Roanoke VA and Richmond Memorial Hospital Richmond, VA, Clinical Manager Memorial Regional Medical Center Mechanicsville, VA
Pharmacometrics Lead

[Name]

Credentials: [Credentials]

Expertise:

  • [Population PK modeling experience]
  • [Statistical methods expertise]
  • [Software tools: NONMEM, Monolix, etc.]

Professional Affiliations

  • Professional organizations - ASHP, VSHP
  • [Research collaborations]

Contact Information

For clinical inquiries: marshallmrmc@gmail.com

For technical support: marshallmrmc@gmail.com

Institution: No Affiliation

9. Technical Specifications

Platform Architecture

  • Framework: Flask 3.1.3 (Python web framework)
  • Database: PostgreSQL 16 (production), SQLite (development)
  • Frontend: Bootstrap 4.6.2, Chart.js 4.4.2, jQuery 3.6.0
  • Deployment: Render.com (PaaS), Waitress WSGI server
  • Authentication: Flask-Login with bcrypt password hashing
  • CSRF Protection: Flask-WTF with session-based tokens

Scientific Computing Libraries

  • SciPy 1.13+: Optimization algorithms (L-BFGS-B, Nelder-Mead), statistical functions
  • NumPy: Numerical array operations, mathematical functions
  • Pandas: Data manipulation and analysis
  • Statsmodels: Multiple linear regression, statistical modeling

Optimization Algorithms

Algorithm Use Case Advantages
L-BFGS-B Individual Bayesian fitting (2 parameters) Fast convergence, bounded optimization, gradient-based
Nelder-Mead Population parameter optimization (3-5 parameters) Derivative-free, robust to local minima, no gradient required

Data Security & Privacy

  • Multi-Tenant Architecture: Site-based patient isolation with indexed queries
  • Access Control: Role-based permissions (user, admin)
  • Password Security: Bcrypt hashing with work factor 12
  • Session Management: Secure server-side sessions with timeout
  • HIPAA Considerations: PHI access restricted by site code; audit trails for data modifications

Performance Optimization

  • Caching: Analytics dashboard caching (3-4s live → <50ms cached)
  • Database Indexing: Composite indexes on frequently queried columns (site_code, is_active, drug, fit_datetime)
  • Query Optimization: Eager loading of relationships, minimized N+1 queries
  • Chart Rendering: Client-side Chart.js for interactive visualizations

Validation & Testing

  • Unit Tests: Core PK calculation functions validated against published examples
  • Statistical Validation: Paired t-tests, Wilcoxon signed-rank tests, Cohen's d effect size
  • Cross-Validation: Population model performance assessed on held-out patient cohorts
  • Clinical Review: Dose recommendations reviewed by clinical pharmacists before implementation

10. Quality Assurance & Validation

Model Validation Metrics

Metric Definition Target Clinical Interpretation
R² (Coefficient of Determination) 1 - (SSresidual / SStotal) ≥0.70 Proportion of variance explained by model
MPE% (Mean Prediction Error) Mean[(Pred - Obs) / Obs] × 100 ±10% Systematic bias; positive = over-prediction
MAPE% (Mean Absolute Prediction Error) Mean[|Pred - Obs| / Obs] × 100 ≤15% Average magnitude of prediction error
RMSE (Root Mean Squared Error) √[Mean((Pred - Obs)²)] Drug-specific Overall prediction precision in concentration units
NRMSE% (Normalized RMSE) (RMSE / Meanobs) × 100 ≤20% Dimensionless error; enables cross-drug comparison

Population Model Optimization Process

  1. Data Collection: Minimum 50-100 patients with complete dosing/level/SCr data
  2. Initial Fit: Fit all patients with published population parameters
  3. Objective Function: Minimize weighted sum of squared errors (10% CV on measurements)
  4. Nelder-Mead Optimization: 500 iterations, bounds set by user (typically ±3 SD)
  5. Statistical Testing: Paired t-test, Wilcoxon test, Cohen's d vs. current parameters
  6. Stratified Analysis: Performance by CKD stage, BMI, age, location
  7. Dosing Impact: Calculate dose changes for target attainment
  8. Clinical Review: Pharmacy committee review before implementation

Ongoing Monitoring

  • Target Attainment: Track percentage achieving therapeutic AUC (vancomycin) or peak/trough (aminoglycosides)
  • Nephrotoxicity Surveillance: Monitor SCr changes ≥0.5 mg/dL or ≥50% increase
  • Dose Appropriateness: Quarterly review of dose-level pairs for guideline compliance
  • Model Drift Detection: Annual re-optimization to detect population parameter changes

Error Handling & Alerts

  • Missing Data Validation: Alerts when required demographics or labs are absent
  • Physiologic Range Checks: Flags values outside expected ranges (e.g., CrCl <5 or >200 mL/min)
  • Extreme Dose Warnings: Highlights doses >2× or <0.5× typical values for review
  • Convergence Failure Notification: Logs optimization failures for manual review

Documentation & Audit Trail

  • All PK fits stored with patient demographics snapshot (age, weight, BSA, CrCl at time of fit)
  • Model type, algorithm, and parameters recorded for each fit
  • Dose recommendations tracked with predicted peak/trough/AUC
  • Population parameter changes logged with date, user, and clinical justification

Medical Disclaimer & Limitations of Liability

FOR USE BY LICENSED HEALTHCARE PROFESSIONALS ONLY

This platform is intended solely for use by licensed pharmacists, physicians, and other qualified healthcare professionals trained in pharmacokinetics and therapeutic drug monitoring. Unauthorized use is prohibited.

No Warranties - Express or Implied

This software, including all recommendations, calculations, equations, population models, and content, is provided "AS IS" and "AS AVAILABLE" without warranty of any kind, either express or implied, including but not limited to:

  • Warranties of merchantability
  • Fitness for a particular purpose
  • Accuracy, completeness, or reliability of information
  • Non-infringement of intellectual property
  • Uninterrupted or error-free operation
  • Freedom from bugs, viruses, or other harmful components
Not a Substitute for Professional Clinical Judgment

This platform is designed as an educational and informational tool to support, but NOT replace, the clinical judgment, experience, training, and expertise of qualified healthcare professionals. Key limitations include:

  • Clinical Decision Responsibility: All treatment decisions, including but not limited to medication selection, dosing, monitoring, and adjustments, remain the sole responsibility of the treating healthcare professional.
  • Independent Verification Required: Users MUST independently verify all dosing recommendations, calculations, patient data, drug concentrations, and clinical parameters before patient administration.
  • Patient-Specific Factors: The software cannot account for all patient-specific variables, comorbidities, drug interactions, genetic factors, or clinical circumstances that may affect drug disposition and response.
  • Model Limitations: Population pharmacokinetic models are based on literature data and may not accurately predict individual patient responses. Bayesian estimates are probabilistic and subject to uncertainty.
  • Data Quality: Accuracy of recommendations depends entirely on the accuracy and completeness of user-entered patient data, laboratory values, and dosing history.
Limitations of Liability

To the maximum extent permitted by applicable law, the developers, contributors, maintainers, distributors, and any affiliated institutions or individuals (collectively "the Providers") shall NOT BE LIABLE for any damages whatsoever arising from or related to the use of, or inability to use, this platform, including but not limited to:

  • Direct Damages: Medication errors, incorrect dosing, adverse drug reactions, treatment failures, or any patient harm
  • Indirect Damages: Loss of data, business interruption, loss of professional reputation, regulatory actions, or legal proceedings
  • Consequential Damages: Patient morbidity or mortality, extended hospitalizations, additional medical interventions, or long-term sequelae
  • Special or Punitive Damages: Even if the Providers have been advised of the possibility of such damages

This limitation applies whether the claim is based on warranty, contract, tort (including negligence), strict liability, or any other legal theory.

User Responsibilities and Acknowledgments

By using this platform, healthcare professionals acknowledge and agree to the following responsibilities:

  1. Licensure & Competence: User is a licensed healthcare professional with appropriate training, credentials, and scope of practice to prescribe, monitor, and adjust antimicrobial therapy.
  2. Data Verification: User will verify accuracy of all patient data, laboratory values, weights, heights, renal function estimates, and dosing history before relying on any recommendations.
  3. Calculation Validation: User will independently validate all pharmacokinetic calculations, dose recommendations, and monitoring parameters using clinical judgment and alternative methods when appropriate.
  4. Institutional Policies: User will follow all applicable institutional policies, protocols, formulary restrictions, antimicrobial stewardship guidelines, and regulatory requirements.
  5. Informed Consent: User will obtain appropriate informed consent from patients or their legal representatives as required by law and institutional policy.
  6. Clinical Monitoring: User will implement appropriate clinical and laboratory monitoring to assess efficacy, toxicity, and need for dosing adjustments.
  7. Specialist Consultation: User will consult appropriate specialists (e.g., infectious disease, nephrology, clinical pharmacology) when clinical complexity exceeds their expertise.
  8. Documentation: User will maintain appropriate clinical documentation of all assessments, recommendations, and rationale for dosing decisions.
  9. Error Reporting: User acknowledges responsibility to report any suspected software errors, calculation discrepancies, or adverse events through appropriate channels.
  10. Assumption of Risk: User assumes all risks associated with use of this platform and agrees to indemnify and hold harmless the Providers from any claims, liabilities, damages, or expenses arising from such use.
Clinical Practice Standards

Use of this platform does not alter or supersede applicable standards of care, professional practice guidelines, or regulatory requirements. Healthcare professionals must:

  • Adhere to current clinical practice guidelines (e.g., ASHP, IDSA, SIDP consensus statements)
  • Comply with institutional formulary restrictions and antimicrobial stewardship policies
  • Follow FDA-approved labeling and dosing recommendations where applicable
  • Maintain appropriate professional liability insurance coverage
  • Engage in continuing education to maintain competency in pharmacokinetics and therapeutic drug monitoring
Software Updates and Modifications

The Providers reserve the right to modify, update, or discontinue this platform at any time without notice. Users are responsible for ensuring they are using the most current version and for reviewing any updates to methodology, models, or calculations.

Governing Law and Jurisdiction

Any disputes arising from use of this platform shall be governed by applicable laws. Users agree that use of this platform constitutes acceptance of these terms and limitations.

Questions or Concerns?

If you have questions about appropriate use of this platform, clinical methodology, or specific patient cases, please consult with experienced colleagues, specialists, or clinical pharmacokinetics experts. Report any suspected software errors or safety concerns to the development team immediately.

Last Updated: July 14, 2026 | Platform Version: 2.0
For questions or feedback, please contact the development team through the main application.