Model-Informed Precision Dosing: Background, Requirements, Validation, Implementation, and Forward Trajectory of Individualizing Drug Therapy.

Model-informed precision dosing (MIPD) has become synonymous with modern approaches for individualizing drug therapy, in which the characteristics of each patient are considered as opposed to applying a one-size-fits-all alternative. This review provides a brief account of the current knowledge, practices, and opinions on MIPD while defining an achievable vision for MIPD in clinical care based on available evidence. We begin with a historical perspective on variability in dose requirements and then discuss technical aspects of MIPD, including the need for clinical decision support tools, practical validation, and implementation of MIPD in health care. We also discuss novel ways to characterize patient variability beyond the common perceptions of genetic control. Finally, we address current debates on MIPD from the perspectives of the new drug development, health economics, and drug regulations.

Estimation of adherence to urate-lowering therapy in people living with gout using Australia's Pharmaceutical Benefits Scheme and patient-reported dosing.

AIMS: The aim of this study was to estimate adherence to urate-lowering therapy (ULT), predominately allopurinol, from Australia's Pharmaceutical Benefits Scheme (PBS) claims database in association with (1) patient-reported doses and (2) World Health Organization's (WHO) defined daily doses (DDD), namely, allopurinol (400 mg/day) or febuxostat (80 mg/day). METHODS: Proportion of days covered (PDC) was calculated in 108 Gout App (Gout APP) trial participants with at least two recorded ULT dispensings in an approximately 12-month period before provision of intervention or control apps. Adherence was defined as PDC ≥80%. We measured the correlation between the two methods of calculating PDC using a Wilcoxon signed rank test. Agreement between ULT-taking status (self-reports) and ULT-dispensed status (PBS records) was tested with Cohen's kappa (κ), and positive and negative percent agreement. RESULTS: Allopurinol was prescribed in 93.5% of participants taking ULT. Their self-reported mean daily dose (SD) was 291 (167) mg/day. Mean PDC (SD) for allopurinol was 83% (21%) calculated using self-reported dose, and 63% (24%) using WHO's DDD. Sixty-three percent of allopurinol users were identified as adherent (PDC ≥80%) using self-reported dose. There was good agreement between self-reported ULT use and PBS dispensing claims (κ = 0.708, P < .001; positive percent agreement = 90%, negative percent agreement = 82%). CONCLUSIONS: Participant-reported allopurinol daily doses, in addition to PBS dispensing claims, may enhance confidence in estimating PDC and adherence compared to using DDD. This approach improves adherence estimations from pharmaceutical claims datasets for medications where daily doses vary between individuals or where there is a wide therapeutic dose range.

The development and evaluation of dose-prediction tools for allopurinol therapy (Easy-Allo tools).

AIMS: Dose escalation at the initiation of allopurinol therapy can be protracted and resource intensive. Tools to predict the allopurinol doses required to achieve target serum urate concentrations would facilitate the implementation of more efficient dose-escalation strategies. The aim of this research was to develop and externally evaluate allopurinol dosing tools, one for use when the pre-urate-lowering therapy serum urate is known (Easy-Allo1) and one for when it is not known (Easy-Allo2). METHODS: A revised population pharmacokinetic-pharmacodynamic model was developed using data from 653 people with gout. Maintenance doses to achieve the serum urate target of <0.36 mmol L-1 in >80% of individuals were simulated and evaluated against external data. The predicted and observed allopurinol doses were compared using the mean prediction error (MPE) and root mean square error (RMSE). The proportion of Easy-Allo predicted doses within 100 mg of the observed was quantified. RESULTS: Allopurinol doses were predicted by total body weight, baseline urate, ethnicity and creatinine clearance. Easy-Allo1 produced unbiased and suitably precise dose predictions (MPE 2 mg day-1 95% confidence interval [CI] -13-17, RMSE 91%, 90% within 100 mg of the observed dose). Easy-Allo2 was positively biased by about 70 mg day-1 and slightly less precise (MPE 70 mg day-1 95% CI 52-88, RMSE 131%, 71% within 100 mg of the observed dose). CONCLUSIONS: The Easy-Allo tools provide a guide to the allopurinol maintenance dose requirement to achieve the serum urate target of <0.36 mmol L-1 and will aid in the development of novel dose-escalation strategies for allopurinol therapy.

The impact of peritoneal dialysis on oxypurinol and urate elimination in people with gout.

Gout affects 15%-30% of individuals with advanced kidney disease. Allopurinol which is rapidly and extensively metabolised to an active metabolite, oxypurinol, is the most commonly prescribed urate-lowering therapy. Oxypurinol is almost entirely eliminated by the kidneys (>95%) and has an elimination half-life of 18-30 h in those with normal kidney function. However, oxypurinol pharmacokinetics are poorly understood in individuals with kidney failure on peritoneal dialysis. This study characterised the elimination of oxypurinol and urate in people with gout receiving peritoneal dialysis. Oxypurinol steady-state oral clearance (CL/F), elimination half-life as well as kidney (CLk) and peritoneal (CLpd) clearances for oxypurinol and urate were calculated from the plasma, urine and dialysate concentration data for each individual. Our results demonstrate that oxypurinol and urate are removed by peritoneal dialysis, accounting for more than 50% of oxypurinol and urate clearances. An allopurinol dose about 50%-60% lower than the usual dose used for a patient with normal kidney function will provide adequate urate-lowering therapy.

The experiences and perspectives of people with gout on urate self-monitoring.

INTRODUCTION: Gout management remains suboptimal despite safe and effective urate-lowering therapy. Self-monitoring of urate may improve gout management, however, the acceptability of urate self-monitoring by people with gout is unknown. The aim of this study was to explore the experiences of urate self-monitoring in people with gout. METHODS: Semistructured interviews were conducted with people taking urate-lowering therapy (N = 30) in a 12-month trial of urate self-monitoring in rural and urban Australia. Interviews covered the experience of monitoring and its effect on gout self-management. Deidentified transcripts were analysed thematically. RESULTS: Participants valued the ability to self-monitor and gain more understanding of urate control compared with the annual monitoring ordered by their doctors. Participants indicated that self-monitoring at home was easy, convenient and informed gout self-management behaviours such as dietary modifications, hydration, exercise and medication routines. Many participants self-monitored to understand urate concentration changes in response to feeling a gout flare was imminent or whether their behaviours, for example, alcohol intake, increased the risk of a gout flare. Urate concentrations were shared with doctors mainly when they were above target to seek management support, and this led to allopurinol dose increases in some cases. CONCLUSION: Urate self-monitoring was viewed by people with gout as convenient and useful for independent management of gout. They believed self-monitoring achieved better gout control with a less restricted lifestyle. Urate data was shared with doctors at the patient's discretion and helped inform clinical decisions, such as allopurinol dose changes. Further research on implementing urate self-monitoring in routine care would enable an evaluation of its impact on medication adherence and clinical outcomes, as well as inform gout management guidelines. PATIENT OR PUBLIC CONTRIBUTION: One person with gout, who was not a participant, was involved in the study design by providing feedback and pilot testing the semistructured interview guide. In response to their feedback, subsequent modifications to the interview guide were made to improve the understandability of the questions from a patient perspective. No additional questions were suggested.

A framework for understanding sources of bias in medication adherence research.

The sources of bias in medication adherence research have not been comprehensively explored. We aimed to identify biases expected to affect adherence research and to develop a framework for mapping these onto the phases of adherence (initiation, implementation and discontinuation). A literature search was conducted, key papers were reviewed and a Catalogue of Bias was consulted. The specific biases related to adherence measurement and metrics were mapped onto the phases of adherence using a tabular matrix. Twenty-three biases were identified, of which 11 were specifically relevant to adherence measures and metrics. The mapping framework showed differences in the numbers and types of biases associated with each measure and metric while highlighting those common to many adherence study designs (e.g., unacceptability bias and apprehension bias). The framework will inform the design of adherence studies and the development of risk of bias tools for adherence research.

An allopurinol adherence tool using plasma oxypurinol concentrations.

AIMS: This study aimed to develop and evaluate an allopurinol adherence tool based on steady-state oxypurinol plasma concentrations, allopurinol's active metabolite. METHODS: Plasma oxypurinol concentrations were simulated stochastically from an oxypurinol pharmacokinetic model for allopurinol doses of 100-800 mg daily, accounting for differences in renal function, diuretic use and ethnicity. For each scenario, the 20th percentile for peak and trough concentrations defined the adherence threshold, below which imperfect adherence was assumed. Predictive performance was evaluated using both simulated low adherence and against data from 146 individuals with paired oxypurinol plasma concentrations and adherence measures. Sensitivity and specificity (S&S), negative and positive predictive values (NPV, PPV) and receiver operating characteristic (ROC) area under the curve (AUC) were determined. The predictive performance of the tool was evaluated using adherence data from an external study (CKD-FIX). RESULTS: The allopurinol adherence tool produced S&S values for trough thresholds of 89-98% and 76-84%, respectively, and 90%-98% and 76-83% for peak thresholds. PPV and NPV were 79-84% and 88-94%, respectively, for trough and 80-85% and 89-98%, respectively, for peak concentrations. The ROC AUC values ranged from 0.84 to 0.88 and from 0.86 to 0.89 for trough and peak concentrations, respectively. S&S values for the external evaluation were found to be 75.8% and 86.5%, respectively, producing an ROC AUC of 0.8113. CONCLUSION: A tool to identify people with gout who require additional support to maintain adherence using plasma oxypurinol concentrations was developed and evaluated. The predictive performance of the tool is suitable for adherence screening in clinical trials and may have utility in some clinical practice settings.

Evaluation of designs for renal drug studies based on the European Medicines Agency and Food and Drug Administration guidelines for drugs that are predominantly secreted.

AIMS: Dose adjustment for drugs eliminated by the kidneys generally assume a linear relationship between renal drug clearance (CLR ) and glomerular filtration rate (GFR). This assumption may not hold for drugs that undergo extensive tubular secretion where nonlinearity in drug handling is expected. The aim of this study is to determine if renal drug study designs recommended by the European Medicines Agency (EMA) and Food and Drug Administration (FDA) could distinguish linear from nonlinear renal drug handling. METHODS: In this simulation and estimation study, the study designs based on the EMA and FDA guidelines for Phase I renal drug studies were evaluated for their ability to discriminate a linear from a nonlinear relationship between CLR and GFR. The number of subjects for each simulated study ranged from 4 to 960. Power, relative standard error and bias were calculated. RESULTS: Study designs under the EMA and FDA guidelines required ≥8 and ≥48 subjects, respectively, to achieve ≥80% power to discriminate a linear from nonlinear relationship between CLR and GFR. The relative standard error of estimated parameters were 13-37 and 17-44% for the designs with 24 subjects under the EMA and FDA guidelines, respectively. The bias in parameter estimates under the EMA designs were not evident, however, they were biased (13-21%) under the FDA designs. CONCLUSION: The EMA design was found to require fewer subjects (n = 8) compared to the FDA (n = 48) to discriminate linear from nonlinear drug renal handling at ≥80% study power while both the designs perform poorly for the parameter precision.

Does the intact nephron hypothesis provide a reasonable model for metformin dosing in chronic kidney disease?

This research explored the intact nephron hypothesis (INH) as a model for metformin dosing in patients with chronic kidney disease (CKD). The INH assumes that glomerular filtration rate (GFR) will account for all kidney drug handling even for drugs eliminated by tubular secretion like metformin. We conducted two studies: (1) a regression analysis to explore the relationship between metformin clearance and eGFR metrics, and (2) a joint population pharmacokinetic analysis to test the relationship between metformin renal clearance and gentamicin clearance. The relationship between metformin renal clearance and eGFR metrics and gentamicin clearance was found to be linear, suggesting that a proportional dose reduction based on GFR in patients with CKD is reasonable.

The pharmacokinetics of metformin in patients receiving intermittent haemodialysis.

The aims of this study were to characterise the population pharmacokinetics of metformin in patients receiving haemodialysis, and to determine the doses that will maintain median metformin plasma concentrations below 5 mg L-1 for a typical individual. Metformin plasma concentrations from 5 patients receiving thrice weekly intermittent haemodialysis followed by metformin 500 mg postdialysis were fitted to a published pharmacokinetic model. Additional models to describe the dialytic pharmacokinetics of metformin were explored. Doses of 250 and 500 postdialysis were simulated from the model for a typical haemodialysis patient. The published 2-compartment pharmacokinetic model with an additional parameter to describe haemodialysis clearance provided a reasonable fit to the data. Deterministic simulations from the model for a typical individual suggest that metformin doses of 250-500 mg postdialysis and 250 mg given once daily should maintain median metformin plasma concentrations below 5 mg L-1 .

The pharmacokinetics of meropenem and piperacillin-tazobactam during sustained low efficiency haemodiafiltration (SLED-HDF).

OBJECTIVE: We analysed the pharmacokinetics of meropenem and piperacillin-tazobactam in patients undergoing a standardised session of sustained low efficiency haemodiafiltration (SLED-HDF) to inform the dosing of these drugs in an acute setting. PARTICIPANTS: Six stable patients with end-stage kidney disease. METHODS: An open-label pilot pharmacokinetic study of meropenem and piperacillin-tazobactam. SLED-HDF was undertaken for 4 h. Plasma drug concentrations were measured pre- and post-filter and in the effluent at multiple time points. The pharmacokinetic data was analysed using non-compartmental methods. The fraction of time that individual plasma concentration profiles were predicted to remain above the MIC break-points for commonly isolated gram-negative pathogens during a prolonged SLED-HDF session was assessed using two targets; fT > MIC (fraction of time above the MIC) and the more aggressive fT > 4 × MIC (fraction of time above 4 × MIC). RESULTS: Meropenem total and SLED-HDF clearance ranged from 141 to 180 mL/min and 126-205 mL/min, respectively. Piperacillin total and SLED-HDF clearance values ranged from 131 to 252 mL/min and 135-162 mL/min, respectively. Our results suggest that prolonged SLED-HDF (12 h) will only maintain a sufficient meropenem and piperacillin-tazobactam plasma concentration to achieve a target of fT > MIC for gram-negative pathogens (MIC 2 mg/L-meropenem, 8 mg/L-piperacillin-tazobactam) for less than 40% of the time. Plasma concentrations would be inadequate to achieve the more aggressive target of 100 % fT > 4xMIC target recommended for critically unwell patients. CONCLUSIONS: The pharmacokinetic data obtained from this pilot study demonstrate significant quantities of meropenem and piperacillin are removed during a SLED-HDF session. This may lead to subtherapeutic concentrations of piperacillin and meropenem over the duration of HDF session. TRIAL REGISTRATION: Australasian Clinical Trials Registry Network (ACTRN12616000078459).

The intact nephron hypothesis as a model for renal drug handling.

PURPOSE: The intact nephron hypothesis (INH) states that impaired renal function results from a reduction in the number of complete (intact) nephrons. Under this model, renal drug clearance is assumed to be a linear function of glomerular filtration while tubular handling is ignored. The aims of this study were to systematically review published studies designed to test the INH and to assess the strength of the study designs used. METHODS: A systematic literature search was conducted in MEDLINE, EMBASE and Google Scholar. Studies specifically designed to understand the relationship between glomerular and tubular function across different levels of renal function were included. Studies that found a linear relationship between GFR and tubular clearance were deemed to support the INH while studies that found a non-linear relationship did not support the INH. Study design was accessed using a bespoke strength of evidence score. RESULTS: Thirty studies met the criteria for inclusion. Of these, 24 did not support the INH. Studies that did not support the INH used methods for measuring tubular clearance that were more robust and included subjects with a wider range of GFR values than studies that supported the INH. DISCUSSION: Our results suggest that the INH may not be a suitable general model for renal drug handling, particularly for drugs that are eliminated by tubular mechanisms. Further studies to assess the clinical importance of a non-linear relationship between drug clearance and GFR are warranted.

The impact of diuretic use and ABCG2 genotype on the predictive performance of a published allopurinol dosing tool.

AIM: This research aims to evaluate the predictive performance of a published allopurinol dosing tool. METHODS: Allopurinol dose predictions were compared to the actual dose required to achieve serum urate (SU) <0.36 mmol l-1 using mean prediction error. The influence of patient factors on dose predictions was explored using multilinear regression. RESULTS: Allopurinol doses were overpredicted by the dosing tool; however, this was minimal in patients without diuretic therapy (MPE 63 mg day-1 , 95% CI 40-87) compared to those receiving diuretics (MPE 295 mg day-1 , 95% CI 260-330, P < 0.0001). ABCG2 genotype (rs2231142, G>T) had an important impact on the dose predictions (MPE 201, 107, 15 mg day-1 for GG, GT and TT, respectively, P < 0.0001). Diuretic use and ABCG2 genotype explained 53% of the variability in prediction error (R2  = 0.53, P = 0.0004). CONCLUSIONS: The dosing tool produced acceptable maintenance dose predictions for patients not taking diuretics. Inclusion of ABCG2 genotype and a revised adjustment for diuretics would further improve the performance of the dosing tool.

'Massive' metformin overdose.

Massive metformin overdose can cause metabolic acidosis with hyperlactatemia. A 55-year-old woman presented 5 h after multidrug overdose, including 132 g extended-release metformin. Continuous venovenous haemodiafiltration (CVVHDF) and noradrenaline were commenced due to metabolic acidosis (pH 7.0, lactate 17 mmol l-1 ) and shock. Despite 3 h of CVVHDF, her acidosis worsened (pH 6.83, lactate 24 mmol l-1 ). Intermittent haemodialysis (IHD) improved acidosis (pH 7.13, lactate 26 mmol l-1 ) but again worsened (pH 6.91, lactate 30 mmol l-1 ) with CVVHDF recommencement. IHD (12 h), CVVHDF (26 h) and vasopressor support for 7 days resulted in survival. Measured metformin concentrations were extremely high with a peak of 292 µg ml-1 at 8 h postingestion. IHD, but not CVVHDF in this case, was associated with improvement in metabolic acidosis and hyperlactataemia. Pharmacokinetic analysis of metformin concentrations found a reduced apparent oral clearance of 8.2 l h-1 and a half-life of approximately 30 h. During IHD, the apparent oral clearance increased to 22.2 l h-1 with an approximate half-life of 10 h. The impact of prolonged oral absorption from a pharmacobezoar and redistribution of metformin from peripheral sites (including erythrocytes) on the pharmacokinetic profile cannot be determined from the data available.

A general empirical model for renal drug handling in pharmacokinetic analyses.

Dose adjustment in renal insufficiency is generally based on the assumption that renal drug clearance is related linearly to glomerular filtration rate. The theory underpinning this model is the intact nephron hypothesis, which says that impaired renal function is caused by a reduction in the number of complete (intact) nephrons. The purpose of the present commentary is to propose a general empirical model for renal drug handling. We will explore models for renal function under two scenarios: one that aligns with the intact nephron hypothesis, and one that relaxes the assumptions of this hypothesis. We propose that a nonlinear, non-intact nephron model will allow for differences in renal drug handling, while incorporating the intact nephron hypothesis model as a special case.

Individualising the dose of allopurinol in patients with gout.

AIMS: The aims of the study were to: 1) determine if a plasma oxypurinol concentration-response relationship or an allopurinol dose-response relationship best predicts the dose requirements of allopurinol in the treatment of gout; and 2) to construct a nomogram for calculating the optimum maintenance dose of allopurinol to achieve target serum urate (SU) concentrations. METHODS: A nonlinear regression analysis was used to examine the plasma oxypurinol concentration- and allopurinol dose-response relationships with serum urate. In 81 patients (205 samples), creatinine clearance (CLCR ), concomitant diuretic use and SU concentrations before (UP ) and during (UT ) treatment were monitored across a range of allopurinol doses (D, 50-700 mg daily). Plasma concentrations of oxypurinol (C) were measured in 47 patients (98 samples). Models (n = 47 patients) and predictions from each relationship were compared using F-tests, r2 values and paired t-tests. The best model was used to construct a nomogram. RESULTS: The final plasma oxypurinol concentration-response relationship (UT = UP - C*(UP - UR )/(ID50 + C), r2  = 0.64) and allopurinol dose-response relationship (UT = UP - D* (UP - UR )/(ID50 + D), r2  = 0.60) did not include CLCR or diuretic use as covariates. There was no difference (P = 0.87) between the predicted SU concentrations derived from the oxypurinol concentration- and allopurinol dose-response relationships. The nomogram constructed using the allopurinol dose-response relationship for all recruited patients (n = 81 patients) required pretreatment SU as the predictor of allopurinol maintenance dose. CONCLUSIONS: Plasma oxypurinol concentrations, CLCR and diuretic status are not required to predict the maintenance dose of allopurinol. Using the nomogram, the maintenance dose of allopurinol estimated to reach target concentrations can be predicted from UP .

Deterministic identifiability of population pharmacokinetic and pharmacokinetic-pharmacodynamic models.

Identifiability is an important component of pharmacokinetic-pharmacodynamic (PKPD) model development. Structural identifiability is concerned with the uniqueness of the model parameters for a set of perfect input-output data and deterministic identifiability with the precision of parameter estimation given imperfect input-output data. We introduce two subcategories of deterministic identifiability, external and internal, and consider factors that distinguish between these forms. We define external deterministic identifiability as a function of externally controllable variables, i.e., the design, and internal deterministic identifiability as a function of the model and its parameter values. The concepts are explored using three common PK and PKPD models, and verified for their precision for the selected set of parameter values under optimal design.