A reduction in between subject variability is not mandatory for selecting a new covariate
Abstract Population pharmacokinetic-pharmacodynamic analysis involves nonlinear hierarchical modelling where the mean response in a population and the variability in response from different sources are studied. It generally consists of two model hierarchies: a model for residual error and a model fo...
Ausführliche Beschreibung
Gespeichert in:
| Autor*in: |
Lagishetty, Chakradhar V. [verfasserIn] Vajjah, Pavan [verfasserIn] Duffull, Stephen B. [verfasserIn] |
|---|
| Format: |
E-Artikel |
|---|---|
| Sprache: |
Englisch |
| Erschienen: |
2012 |
|---|
| Schlagwörter: |
Non-nested and nested covariate models Predictable between subject variability Random between subject variability and population pharmacokinetics |
|---|
| Übergeordnetes Werk: |
Enthalten in: Journal of Pharmacokinetics and Biopharmaceutics - Kluwer Academic Publishers-Plenum Publishers, 1973, 39(2012), 4 vom: 06. Juli, Seite 383-392 |
|---|---|
| Übergeordnetes Werk: |
volume:39 ; year:2012 ; number:4 ; day:06 ; month:07 ; pages:383-392 |
| Links: |
|---|
| DOI / URN: |
10.1007/s10928-012-9256-2 |
|---|
| Katalog-ID: |
SPR014710153 |
|---|
| LEADER | 01000caa a22002652 4500 | ||
|---|---|---|---|
| 001 | SPR014710153 | ||
| 003 | DE-627 | ||
| 005 | 20230519141606.0 | ||
| 007 | cr uuu---uuuuu | ||
| 008 | 201006s2012 xx |||||o 00| ||eng c | ||
| 024 | 7 | |a 10.1007/s10928-012-9256-2 |2 doi | |
| 035 | |a (DE-627)SPR014710153 | ||
| 035 | |a (SPR)s10928-012-9256-2-e | ||
| 040 | |a DE-627 |b ger |c DE-627 |e rakwb | ||
| 041 | |a eng | ||
| 100 | 1 | |a Lagishetty, Chakradhar V. |e verfasserin |4 aut | |
| 245 | 1 | 2 | |a A reduction in between subject variability is not mandatory for selecting a new covariate |
| 264 | 1 | |c 2012 | |
| 336 | |a Text |b txt |2 rdacontent | ||
| 337 | |a Computermedien |b c |2 rdamedia | ||
| 338 | |a Online-Ressource |b cr |2 rdacarrier | ||
| 520 | |a Abstract Population pharmacokinetic-pharmacodynamic analysis involves nonlinear hierarchical modelling where the mean response in a population and the variability in response from different sources are studied. It generally consists of two model hierarchies: a model for residual error and a model for heterogeneity termed between subject variance (BSV). The overall variability in a parameter within a population termed population parameter variance (PPV) consists of within subject variance (WSV) and BSV. Both these variances can further be split into random and predictable components. The predictable component of BSV (termed BSVP) is explained by covariates, individual characteristics e.g. weight. As BSVP increases, the remaining unpredictable (or random) between subject variability (BSVR) decreases since BSV = BSVP + BSVR, and BSV is a constant in any given data set. Since BSV and BSVR are estimated from the base and full covariate models, respectively, then BSVP = BSV−BSVR. The aim of this study was to explore the hypothesis, that a significant covariate may not always decrease BSVR. The specific aims were: (1) to explore circumstances where BSVR may not be reduced when adding a significantly correlated covariate and (2) to explore whether specific models for covariates may eliminate this anomaly when assessing BSVR. Simulations were performed using MATLAB (2011a) and estimation using NONMEM (ver 7.2) with FOCE and INTERACTION. A 1-compartment intravenous bolus PK model was used for simulation following a single unit dose (d = 1). The BSV of clearance [BSV(CL)] was described according to a log-normal distribution model with mean zero and variance %$ \omega^{ 2} %$. An additive random unexplained variability was assumed. Initially, we show through a simple simulation that BSVR can increase when a significantly correlated covariate is added to the model. We follow this with five simulation scenarios, A to E, that have various levels of correlation between the continuous covariate (Z) and CL ranging from 0 to 100 %. Each simulated scenario was replicated 100 times and estimated by a base model (i.e. without covariate addition) and six covariate models (M1–M6) which included non-nested (M1), nested (M2), and two types of interaction models for each of M1 and M2; non-nested interaction (M3, M5), nested interaction (M4, M6). Initially, through a motivating example we show that BSVR may not reduce even when there is 50 % correlation between the covariate Z and CL. It was found that with 0 % correlation M1, the non-nested covariate model (NNCM) resulted in negative BSVP (inflated BSVR) whereas M2, the nested covariate model (NCM), resulted in a calculated BSVP of zero. NNCM (M1) shows negative BSVP (BSVR > BSV) with correlation as high as 50 % and this model needs a minimum of 75 % correlation to show a positive BSVP. NCM (M2) shows positive but downwardly biased BSVP with 25, 50 and 75 % correlations. However, inclusion of a covariate–eta interaction term for both types of covariate models resulted in greater BSVP for 25, 50 and 75 % correlation scenarios compared to NNCM and NCM respectively. For 100 % correlation, it was found that covariate–eta interaction models show the same BSVP as the models without the interaction term, i.e. under perfect positive correlation all models perform similarly and correctly. It was found that a significantly correlated covariate may not reduce BSVR and in fact it may inflate the BSVR due to statistical misspecification of the covariate model. Incorporating statistical models that account for the covariate–eta interaction may be useful diagnostically in identifying the variability explained by covariates. | ||
| 650 | 4 | |a Non-nested and nested covariate models |7 (dpeaa)DE-He213 | |
| 650 | 4 | |a Predictable between subject variability |7 (dpeaa)DE-He213 | |
| 650 | 4 | |a Random between subject variability and population pharmacokinetics |7 (dpeaa)DE-He213 | |
| 700 | 1 | |a Vajjah, Pavan |e verfasserin |4 aut | |
| 700 | 1 | |a Duffull, Stephen B. |e verfasserin |4 aut | |
| 773 | 0 | 8 | |i Enthalten in |t Journal of Pharmacokinetics and Biopharmaceutics |d Kluwer Academic Publishers-Plenum Publishers, 1973 |g 39(2012), 4 vom: 06. Juli, Seite 383-392 |w (DE-627)SPR014694166 |7 nnns |
| 773 | 1 | 8 | |g volume:39 |g year:2012 |g number:4 |g day:06 |g month:07 |g pages:383-392 |
| 856 | 4 | 0 | |u https://dx.doi.org/10.1007/s10928-012-9256-2 |z lizenzpflichtig |3 Volltext |
| 912 | |a GBV_USEFLAG_A | ||
| 912 | |a SYSFLAG_A | ||
| 912 | |a GBV_SPRINGER | ||
| 912 | |a SSG-OLC-PHA | ||
| 912 | |a GBV_ILN_40 | ||
| 951 | |a AR | ||
| 952 | |d 39 |j 2012 |e 4 |b 06 |c 07 |h 383-392 | ||
| author_variant |
c v l cv cvl p v pv s b d sb sbd |
|---|---|
| matchkey_str |
lagishettychakradharvvajjahpavanduffulls:2012----:rdcinnewesbetaibltintadtrfre |
| hierarchy_sort_str |
2012 |
| publishDate |
2012 |
| allfields |
10.1007/s10928-012-9256-2 doi (DE-627)SPR014710153 (SPR)s10928-012-9256-2-e DE-627 ger DE-627 rakwb eng Lagishetty, Chakradhar V. verfasserin aut A reduction in between subject variability is not mandatory for selecting a new covariate 2012 Text txt rdacontent Computermedien c rdamedia Online-Ressource cr rdacarrier Abstract Population pharmacokinetic-pharmacodynamic analysis involves nonlinear hierarchical modelling where the mean response in a population and the variability in response from different sources are studied. It generally consists of two model hierarchies: a model for residual error and a model for heterogeneity termed between subject variance (BSV). The overall variability in a parameter within a population termed population parameter variance (PPV) consists of within subject variance (WSV) and BSV. Both these variances can further be split into random and predictable components. The predictable component of BSV (termed BSVP) is explained by covariates, individual characteristics e.g. weight. As BSVP increases, the remaining unpredictable (or random) between subject variability (BSVR) decreases since BSV = BSVP + BSVR, and BSV is a constant in any given data set. Since BSV and BSVR are estimated from the base and full covariate models, respectively, then BSVP = BSV−BSVR. The aim of this study was to explore the hypothesis, that a significant covariate may not always decrease BSVR. The specific aims were: (1) to explore circumstances where BSVR may not be reduced when adding a significantly correlated covariate and (2) to explore whether specific models for covariates may eliminate this anomaly when assessing BSVR. Simulations were performed using MATLAB (2011a) and estimation using NONMEM (ver 7.2) with FOCE and INTERACTION. A 1-compartment intravenous bolus PK model was used for simulation following a single unit dose (d = 1). The BSV of clearance [BSV(CL)] was described according to a log-normal distribution model with mean zero and variance %$ \omega^{ 2} %$. An additive random unexplained variability was assumed. Initially, we show through a simple simulation that BSVR can increase when a significantly correlated covariate is added to the model. We follow this with five simulation scenarios, A to E, that have various levels of correlation between the continuous covariate (Z) and CL ranging from 0 to 100 %. Each simulated scenario was replicated 100 times and estimated by a base model (i.e. without covariate addition) and six covariate models (M1–M6) which included non-nested (M1), nested (M2), and two types of interaction models for each of M1 and M2; non-nested interaction (M3, M5), nested interaction (M4, M6). Initially, through a motivating example we show that BSVR may not reduce even when there is 50 % correlation between the covariate Z and CL. It was found that with 0 % correlation M1, the non-nested covariate model (NNCM) resulted in negative BSVP (inflated BSVR) whereas M2, the nested covariate model (NCM), resulted in a calculated BSVP of zero. NNCM (M1) shows negative BSVP (BSVR > BSV) with correlation as high as 50 % and this model needs a minimum of 75 % correlation to show a positive BSVP. NCM (M2) shows positive but downwardly biased BSVP with 25, 50 and 75 % correlations. However, inclusion of a covariate–eta interaction term for both types of covariate models resulted in greater BSVP for 25, 50 and 75 % correlation scenarios compared to NNCM and NCM respectively. For 100 % correlation, it was found that covariate–eta interaction models show the same BSVP as the models without the interaction term, i.e. under perfect positive correlation all models perform similarly and correctly. It was found that a significantly correlated covariate may not reduce BSVR and in fact it may inflate the BSVR due to statistical misspecification of the covariate model. Incorporating statistical models that account for the covariate–eta interaction may be useful diagnostically in identifying the variability explained by covariates. Non-nested and nested covariate models (dpeaa)DE-He213 Predictable between subject variability (dpeaa)DE-He213 Random between subject variability and population pharmacokinetics (dpeaa)DE-He213 Vajjah, Pavan verfasserin aut Duffull, Stephen B. verfasserin aut Enthalten in Journal of Pharmacokinetics and Biopharmaceutics Kluwer Academic Publishers-Plenum Publishers, 1973 39(2012), 4 vom: 06. Juli, Seite 383-392 (DE-627)SPR014694166 nnns volume:39 year:2012 number:4 day:06 month:07 pages:383-392 https://dx.doi.org/10.1007/s10928-012-9256-2 lizenzpflichtig Volltext GBV_USEFLAG_A SYSFLAG_A GBV_SPRINGER SSG-OLC-PHA GBV_ILN_40 AR 39 2012 4 06 07 383-392 |
| spelling |
10.1007/s10928-012-9256-2 doi (DE-627)SPR014710153 (SPR)s10928-012-9256-2-e DE-627 ger DE-627 rakwb eng Lagishetty, Chakradhar V. verfasserin aut A reduction in between subject variability is not mandatory for selecting a new covariate 2012 Text txt rdacontent Computermedien c rdamedia Online-Ressource cr rdacarrier Abstract Population pharmacokinetic-pharmacodynamic analysis involves nonlinear hierarchical modelling where the mean response in a population and the variability in response from different sources are studied. It generally consists of two model hierarchies: a model for residual error and a model for heterogeneity termed between subject variance (BSV). The overall variability in a parameter within a population termed population parameter variance (PPV) consists of within subject variance (WSV) and BSV. Both these variances can further be split into random and predictable components. The predictable component of BSV (termed BSVP) is explained by covariates, individual characteristics e.g. weight. As BSVP increases, the remaining unpredictable (or random) between subject variability (BSVR) decreases since BSV = BSVP + BSVR, and BSV is a constant in any given data set. Since BSV and BSVR are estimated from the base and full covariate models, respectively, then BSVP = BSV−BSVR. The aim of this study was to explore the hypothesis, that a significant covariate may not always decrease BSVR. The specific aims were: (1) to explore circumstances where BSVR may not be reduced when adding a significantly correlated covariate and (2) to explore whether specific models for covariates may eliminate this anomaly when assessing BSVR. Simulations were performed using MATLAB (2011a) and estimation using NONMEM (ver 7.2) with FOCE and INTERACTION. A 1-compartment intravenous bolus PK model was used for simulation following a single unit dose (d = 1). The BSV of clearance [BSV(CL)] was described according to a log-normal distribution model with mean zero and variance %$ \omega^{ 2} %$. An additive random unexplained variability was assumed. Initially, we show through a simple simulation that BSVR can increase when a significantly correlated covariate is added to the model. We follow this with five simulation scenarios, A to E, that have various levels of correlation between the continuous covariate (Z) and CL ranging from 0 to 100 %. Each simulated scenario was replicated 100 times and estimated by a base model (i.e. without covariate addition) and six covariate models (M1–M6) which included non-nested (M1), nested (M2), and two types of interaction models for each of M1 and M2; non-nested interaction (M3, M5), nested interaction (M4, M6). Initially, through a motivating example we show that BSVR may not reduce even when there is 50 % correlation between the covariate Z and CL. It was found that with 0 % correlation M1, the non-nested covariate model (NNCM) resulted in negative BSVP (inflated BSVR) whereas M2, the nested covariate model (NCM), resulted in a calculated BSVP of zero. NNCM (M1) shows negative BSVP (BSVR > BSV) with correlation as high as 50 % and this model needs a minimum of 75 % correlation to show a positive BSVP. NCM (M2) shows positive but downwardly biased BSVP with 25, 50 and 75 % correlations. However, inclusion of a covariate–eta interaction term for both types of covariate models resulted in greater BSVP for 25, 50 and 75 % correlation scenarios compared to NNCM and NCM respectively. For 100 % correlation, it was found that covariate–eta interaction models show the same BSVP as the models without the interaction term, i.e. under perfect positive correlation all models perform similarly and correctly. It was found that a significantly correlated covariate may not reduce BSVR and in fact it may inflate the BSVR due to statistical misspecification of the covariate model. Incorporating statistical models that account for the covariate–eta interaction may be useful diagnostically in identifying the variability explained by covariates. Non-nested and nested covariate models (dpeaa)DE-He213 Predictable between subject variability (dpeaa)DE-He213 Random between subject variability and population pharmacokinetics (dpeaa)DE-He213 Vajjah, Pavan verfasserin aut Duffull, Stephen B. verfasserin aut Enthalten in Journal of Pharmacokinetics and Biopharmaceutics Kluwer Academic Publishers-Plenum Publishers, 1973 39(2012), 4 vom: 06. Juli, Seite 383-392 (DE-627)SPR014694166 nnns volume:39 year:2012 number:4 day:06 month:07 pages:383-392 https://dx.doi.org/10.1007/s10928-012-9256-2 lizenzpflichtig Volltext GBV_USEFLAG_A SYSFLAG_A GBV_SPRINGER SSG-OLC-PHA GBV_ILN_40 AR 39 2012 4 06 07 383-392 |
| allfields_unstemmed |
10.1007/s10928-012-9256-2 doi (DE-627)SPR014710153 (SPR)s10928-012-9256-2-e DE-627 ger DE-627 rakwb eng Lagishetty, Chakradhar V. verfasserin aut A reduction in between subject variability is not mandatory for selecting a new covariate 2012 Text txt rdacontent Computermedien c rdamedia Online-Ressource cr rdacarrier Abstract Population pharmacokinetic-pharmacodynamic analysis involves nonlinear hierarchical modelling where the mean response in a population and the variability in response from different sources are studied. It generally consists of two model hierarchies: a model for residual error and a model for heterogeneity termed between subject variance (BSV). The overall variability in a parameter within a population termed population parameter variance (PPV) consists of within subject variance (WSV) and BSV. Both these variances can further be split into random and predictable components. The predictable component of BSV (termed BSVP) is explained by covariates, individual characteristics e.g. weight. As BSVP increases, the remaining unpredictable (or random) between subject variability (BSVR) decreases since BSV = BSVP + BSVR, and BSV is a constant in any given data set. Since BSV and BSVR are estimated from the base and full covariate models, respectively, then BSVP = BSV−BSVR. The aim of this study was to explore the hypothesis, that a significant covariate may not always decrease BSVR. The specific aims were: (1) to explore circumstances where BSVR may not be reduced when adding a significantly correlated covariate and (2) to explore whether specific models for covariates may eliminate this anomaly when assessing BSVR. Simulations were performed using MATLAB (2011a) and estimation using NONMEM (ver 7.2) with FOCE and INTERACTION. A 1-compartment intravenous bolus PK model was used for simulation following a single unit dose (d = 1). The BSV of clearance [BSV(CL)] was described according to a log-normal distribution model with mean zero and variance %$ \omega^{ 2} %$. An additive random unexplained variability was assumed. Initially, we show through a simple simulation that BSVR can increase when a significantly correlated covariate is added to the model. We follow this with five simulation scenarios, A to E, that have various levels of correlation between the continuous covariate (Z) and CL ranging from 0 to 100 %. Each simulated scenario was replicated 100 times and estimated by a base model (i.e. without covariate addition) and six covariate models (M1–M6) which included non-nested (M1), nested (M2), and two types of interaction models for each of M1 and M2; non-nested interaction (M3, M5), nested interaction (M4, M6). Initially, through a motivating example we show that BSVR may not reduce even when there is 50 % correlation between the covariate Z and CL. It was found that with 0 % correlation M1, the non-nested covariate model (NNCM) resulted in negative BSVP (inflated BSVR) whereas M2, the nested covariate model (NCM), resulted in a calculated BSVP of zero. NNCM (M1) shows negative BSVP (BSVR > BSV) with correlation as high as 50 % and this model needs a minimum of 75 % correlation to show a positive BSVP. NCM (M2) shows positive but downwardly biased BSVP with 25, 50 and 75 % correlations. However, inclusion of a covariate–eta interaction term for both types of covariate models resulted in greater BSVP for 25, 50 and 75 % correlation scenarios compared to NNCM and NCM respectively. For 100 % correlation, it was found that covariate–eta interaction models show the same BSVP as the models without the interaction term, i.e. under perfect positive correlation all models perform similarly and correctly. It was found that a significantly correlated covariate may not reduce BSVR and in fact it may inflate the BSVR due to statistical misspecification of the covariate model. Incorporating statistical models that account for the covariate–eta interaction may be useful diagnostically in identifying the variability explained by covariates. Non-nested and nested covariate models (dpeaa)DE-He213 Predictable between subject variability (dpeaa)DE-He213 Random between subject variability and population pharmacokinetics (dpeaa)DE-He213 Vajjah, Pavan verfasserin aut Duffull, Stephen B. verfasserin aut Enthalten in Journal of Pharmacokinetics and Biopharmaceutics Kluwer Academic Publishers-Plenum Publishers, 1973 39(2012), 4 vom: 06. Juli, Seite 383-392 (DE-627)SPR014694166 nnns volume:39 year:2012 number:4 day:06 month:07 pages:383-392 https://dx.doi.org/10.1007/s10928-012-9256-2 lizenzpflichtig Volltext GBV_USEFLAG_A SYSFLAG_A GBV_SPRINGER SSG-OLC-PHA GBV_ILN_40 AR 39 2012 4 06 07 383-392 |
| allfieldsGer |
10.1007/s10928-012-9256-2 doi (DE-627)SPR014710153 (SPR)s10928-012-9256-2-e DE-627 ger DE-627 rakwb eng Lagishetty, Chakradhar V. verfasserin aut A reduction in between subject variability is not mandatory for selecting a new covariate 2012 Text txt rdacontent Computermedien c rdamedia Online-Ressource cr rdacarrier Abstract Population pharmacokinetic-pharmacodynamic analysis involves nonlinear hierarchical modelling where the mean response in a population and the variability in response from different sources are studied. It generally consists of two model hierarchies: a model for residual error and a model for heterogeneity termed between subject variance (BSV). The overall variability in a parameter within a population termed population parameter variance (PPV) consists of within subject variance (WSV) and BSV. Both these variances can further be split into random and predictable components. The predictable component of BSV (termed BSVP) is explained by covariates, individual characteristics e.g. weight. As BSVP increases, the remaining unpredictable (or random) between subject variability (BSVR) decreases since BSV = BSVP + BSVR, and BSV is a constant in any given data set. Since BSV and BSVR are estimated from the base and full covariate models, respectively, then BSVP = BSV−BSVR. The aim of this study was to explore the hypothesis, that a significant covariate may not always decrease BSVR. The specific aims were: (1) to explore circumstances where BSVR may not be reduced when adding a significantly correlated covariate and (2) to explore whether specific models for covariates may eliminate this anomaly when assessing BSVR. Simulations were performed using MATLAB (2011a) and estimation using NONMEM (ver 7.2) with FOCE and INTERACTION. A 1-compartment intravenous bolus PK model was used for simulation following a single unit dose (d = 1). The BSV of clearance [BSV(CL)] was described according to a log-normal distribution model with mean zero and variance %$ \omega^{ 2} %$. An additive random unexplained variability was assumed. Initially, we show through a simple simulation that BSVR can increase when a significantly correlated covariate is added to the model. We follow this with five simulation scenarios, A to E, that have various levels of correlation between the continuous covariate (Z) and CL ranging from 0 to 100 %. Each simulated scenario was replicated 100 times and estimated by a base model (i.e. without covariate addition) and six covariate models (M1–M6) which included non-nested (M1), nested (M2), and two types of interaction models for each of M1 and M2; non-nested interaction (M3, M5), nested interaction (M4, M6). Initially, through a motivating example we show that BSVR may not reduce even when there is 50 % correlation between the covariate Z and CL. It was found that with 0 % correlation M1, the non-nested covariate model (NNCM) resulted in negative BSVP (inflated BSVR) whereas M2, the nested covariate model (NCM), resulted in a calculated BSVP of zero. NNCM (M1) shows negative BSVP (BSVR > BSV) with correlation as high as 50 % and this model needs a minimum of 75 % correlation to show a positive BSVP. NCM (M2) shows positive but downwardly biased BSVP with 25, 50 and 75 % correlations. However, inclusion of a covariate–eta interaction term for both types of covariate models resulted in greater BSVP for 25, 50 and 75 % correlation scenarios compared to NNCM and NCM respectively. For 100 % correlation, it was found that covariate–eta interaction models show the same BSVP as the models without the interaction term, i.e. under perfect positive correlation all models perform similarly and correctly. It was found that a significantly correlated covariate may not reduce BSVR and in fact it may inflate the BSVR due to statistical misspecification of the covariate model. Incorporating statistical models that account for the covariate–eta interaction may be useful diagnostically in identifying the variability explained by covariates. Non-nested and nested covariate models (dpeaa)DE-He213 Predictable between subject variability (dpeaa)DE-He213 Random between subject variability and population pharmacokinetics (dpeaa)DE-He213 Vajjah, Pavan verfasserin aut Duffull, Stephen B. verfasserin aut Enthalten in Journal of Pharmacokinetics and Biopharmaceutics Kluwer Academic Publishers-Plenum Publishers, 1973 39(2012), 4 vom: 06. Juli, Seite 383-392 (DE-627)SPR014694166 nnns volume:39 year:2012 number:4 day:06 month:07 pages:383-392 https://dx.doi.org/10.1007/s10928-012-9256-2 lizenzpflichtig Volltext GBV_USEFLAG_A SYSFLAG_A GBV_SPRINGER SSG-OLC-PHA GBV_ILN_40 AR 39 2012 4 06 07 383-392 |
| allfieldsSound |
10.1007/s10928-012-9256-2 doi (DE-627)SPR014710153 (SPR)s10928-012-9256-2-e DE-627 ger DE-627 rakwb eng Lagishetty, Chakradhar V. verfasserin aut A reduction in between subject variability is not mandatory for selecting a new covariate 2012 Text txt rdacontent Computermedien c rdamedia Online-Ressource cr rdacarrier Abstract Population pharmacokinetic-pharmacodynamic analysis involves nonlinear hierarchical modelling where the mean response in a population and the variability in response from different sources are studied. It generally consists of two model hierarchies: a model for residual error and a model for heterogeneity termed between subject variance (BSV). The overall variability in a parameter within a population termed population parameter variance (PPV) consists of within subject variance (WSV) and BSV. Both these variances can further be split into random and predictable components. The predictable component of BSV (termed BSVP) is explained by covariates, individual characteristics e.g. weight. As BSVP increases, the remaining unpredictable (or random) between subject variability (BSVR) decreases since BSV = BSVP + BSVR, and BSV is a constant in any given data set. Since BSV and BSVR are estimated from the base and full covariate models, respectively, then BSVP = BSV−BSVR. The aim of this study was to explore the hypothesis, that a significant covariate may not always decrease BSVR. The specific aims were: (1) to explore circumstances where BSVR may not be reduced when adding a significantly correlated covariate and (2) to explore whether specific models for covariates may eliminate this anomaly when assessing BSVR. Simulations were performed using MATLAB (2011a) and estimation using NONMEM (ver 7.2) with FOCE and INTERACTION. A 1-compartment intravenous bolus PK model was used for simulation following a single unit dose (d = 1). The BSV of clearance [BSV(CL)] was described according to a log-normal distribution model with mean zero and variance %$ \omega^{ 2} %$. An additive random unexplained variability was assumed. Initially, we show through a simple simulation that BSVR can increase when a significantly correlated covariate is added to the model. We follow this with five simulation scenarios, A to E, that have various levels of correlation between the continuous covariate (Z) and CL ranging from 0 to 100 %. Each simulated scenario was replicated 100 times and estimated by a base model (i.e. without covariate addition) and six covariate models (M1–M6) which included non-nested (M1), nested (M2), and two types of interaction models for each of M1 and M2; non-nested interaction (M3, M5), nested interaction (M4, M6). Initially, through a motivating example we show that BSVR may not reduce even when there is 50 % correlation between the covariate Z and CL. It was found that with 0 % correlation M1, the non-nested covariate model (NNCM) resulted in negative BSVP (inflated BSVR) whereas M2, the nested covariate model (NCM), resulted in a calculated BSVP of zero. NNCM (M1) shows negative BSVP (BSVR > BSV) with correlation as high as 50 % and this model needs a minimum of 75 % correlation to show a positive BSVP. NCM (M2) shows positive but downwardly biased BSVP with 25, 50 and 75 % correlations. However, inclusion of a covariate–eta interaction term for both types of covariate models resulted in greater BSVP for 25, 50 and 75 % correlation scenarios compared to NNCM and NCM respectively. For 100 % correlation, it was found that covariate–eta interaction models show the same BSVP as the models without the interaction term, i.e. under perfect positive correlation all models perform similarly and correctly. It was found that a significantly correlated covariate may not reduce BSVR and in fact it may inflate the BSVR due to statistical misspecification of the covariate model. Incorporating statistical models that account for the covariate–eta interaction may be useful diagnostically in identifying the variability explained by covariates. Non-nested and nested covariate models (dpeaa)DE-He213 Predictable between subject variability (dpeaa)DE-He213 Random between subject variability and population pharmacokinetics (dpeaa)DE-He213 Vajjah, Pavan verfasserin aut Duffull, Stephen B. verfasserin aut Enthalten in Journal of Pharmacokinetics and Biopharmaceutics Kluwer Academic Publishers-Plenum Publishers, 1973 39(2012), 4 vom: 06. Juli, Seite 383-392 (DE-627)SPR014694166 nnns volume:39 year:2012 number:4 day:06 month:07 pages:383-392 https://dx.doi.org/10.1007/s10928-012-9256-2 lizenzpflichtig Volltext GBV_USEFLAG_A SYSFLAG_A GBV_SPRINGER SSG-OLC-PHA GBV_ILN_40 AR 39 2012 4 06 07 383-392 |
| language |
English |
| source |
Enthalten in Journal of Pharmacokinetics and Biopharmaceutics 39(2012), 4 vom: 06. Juli, Seite 383-392 volume:39 year:2012 number:4 day:06 month:07 pages:383-392 |
| sourceStr |
Enthalten in Journal of Pharmacokinetics and Biopharmaceutics 39(2012), 4 vom: 06. Juli, Seite 383-392 volume:39 year:2012 number:4 day:06 month:07 pages:383-392 |
| format_phy_str_mv |
Article |
| institution |
findex.gbv.de |
| topic_facet |
Non-nested and nested covariate models Predictable between subject variability Random between subject variability and population pharmacokinetics |
| isfreeaccess_bool |
false |
| container_title |
Journal of Pharmacokinetics and Biopharmaceutics |
| authorswithroles_txt_mv |
Lagishetty, Chakradhar V. @@aut@@ Vajjah, Pavan @@aut@@ Duffull, Stephen B. @@aut@@ |
| publishDateDaySort_date |
2012-07-06T00:00:00Z |
| hierarchy_top_id |
SPR014694166 |
| id |
SPR014710153 |
| language_de |
englisch |
| fullrecord |
<?xml version="1.0" encoding="UTF-8"?><collection xmlns="http://www.loc.gov/MARC21/slim"><record><leader>01000caa a22002652 4500</leader><controlfield tag="001">SPR014710153</controlfield><controlfield tag="003">DE-627</controlfield><controlfield tag="005">20230519141606.0</controlfield><controlfield tag="007">cr uuu---uuuuu</controlfield><controlfield tag="008">201006s2012 xx |||||o 00| ||eng c</controlfield><datafield tag="024" ind1="7" ind2=" "><subfield code="a">10.1007/s10928-012-9256-2</subfield><subfield code="2">doi</subfield></datafield><datafield tag="035" ind1=" " ind2=" "><subfield code="a">(DE-627)SPR014710153</subfield></datafield><datafield tag="035" ind1=" " ind2=" "><subfield code="a">(SPR)s10928-012-9256-2-e</subfield></datafield><datafield tag="040" ind1=" " ind2=" "><subfield code="a">DE-627</subfield><subfield code="b">ger</subfield><subfield code="c">DE-627</subfield><subfield code="e">rakwb</subfield></datafield><datafield tag="041" ind1=" " ind2=" "><subfield code="a">eng</subfield></datafield><datafield tag="100" ind1="1" ind2=" "><subfield code="a">Lagishetty, Chakradhar V.</subfield><subfield code="e">verfasserin</subfield><subfield code="4">aut</subfield></datafield><datafield tag="245" ind1="1" ind2="2"><subfield code="a">A reduction in between subject variability is not mandatory for selecting a new covariate</subfield></datafield><datafield tag="264" ind1=" " ind2="1"><subfield code="c">2012</subfield></datafield><datafield tag="336" ind1=" " ind2=" "><subfield code="a">Text</subfield><subfield code="b">txt</subfield><subfield code="2">rdacontent</subfield></datafield><datafield tag="337" ind1=" " ind2=" "><subfield code="a">Computermedien</subfield><subfield code="b">c</subfield><subfield code="2">rdamedia</subfield></datafield><datafield tag="338" ind1=" " ind2=" "><subfield code="a">Online-Ressource</subfield><subfield code="b">cr</subfield><subfield code="2">rdacarrier</subfield></datafield><datafield tag="520" ind1=" " ind2=" "><subfield code="a">Abstract Population pharmacokinetic-pharmacodynamic analysis involves nonlinear hierarchical modelling where the mean response in a population and the variability in response from different sources are studied. It generally consists of two model hierarchies: a model for residual error and a model for heterogeneity termed between subject variance (BSV). The overall variability in a parameter within a population termed population parameter variance (PPV) consists of within subject variance (WSV) and BSV. Both these variances can further be split into random and predictable components. The predictable component of BSV (termed BSVP) is explained by covariates, individual characteristics e.g. weight. As BSVP increases, the remaining unpredictable (or random) between subject variability (BSVR) decreases since BSV = BSVP + BSVR, and BSV is a constant in any given data set. Since BSV and BSVR are estimated from the base and full covariate models, respectively, then BSVP = BSV−BSVR. The aim of this study was to explore the hypothesis, that a significant covariate may not always decrease BSVR. The specific aims were: (1) to explore circumstances where BSVR may not be reduced when adding a significantly correlated covariate and (2) to explore whether specific models for covariates may eliminate this anomaly when assessing BSVR. Simulations were performed using MATLAB (2011a) and estimation using NONMEM (ver 7.2) with FOCE and INTERACTION. A 1-compartment intravenous bolus PK model was used for simulation following a single unit dose (d = 1). The BSV of clearance [BSV(CL)] was described according to a log-normal distribution model with mean zero and variance %$ \omega^{ 2} %$. An additive random unexplained variability was assumed. Initially, we show through a simple simulation that BSVR can increase when a significantly correlated covariate is added to the model. We follow this with five simulation scenarios, A to E, that have various levels of correlation between the continuous covariate (Z) and CL ranging from 0 to 100 %. Each simulated scenario was replicated 100 times and estimated by a base model (i.e. without covariate addition) and six covariate models (M1–M6) which included non-nested (M1), nested (M2), and two types of interaction models for each of M1 and M2; non-nested interaction (M3, M5), nested interaction (M4, M6). Initially, through a motivating example we show that BSVR may not reduce even when there is 50 % correlation between the covariate Z and CL. It was found that with 0 % correlation M1, the non-nested covariate model (NNCM) resulted in negative BSVP (inflated BSVR) whereas M2, the nested covariate model (NCM), resulted in a calculated BSVP of zero. NNCM (M1) shows negative BSVP (BSVR > BSV) with correlation as high as 50 % and this model needs a minimum of 75 % correlation to show a positive BSVP. NCM (M2) shows positive but downwardly biased BSVP with 25, 50 and 75 % correlations. However, inclusion of a covariate–eta interaction term for both types of covariate models resulted in greater BSVP for 25, 50 and 75 % correlation scenarios compared to NNCM and NCM respectively. For 100 % correlation, it was found that covariate–eta interaction models show the same BSVP as the models without the interaction term, i.e. under perfect positive correlation all models perform similarly and correctly. It was found that a significantly correlated covariate may not reduce BSVR and in fact it may inflate the BSVR due to statistical misspecification of the covariate model. Incorporating statistical models that account for the covariate–eta interaction may be useful diagnostically in identifying the variability explained by covariates.</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Non-nested and nested covariate models</subfield><subfield code="7">(dpeaa)DE-He213</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Predictable between subject variability</subfield><subfield code="7">(dpeaa)DE-He213</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Random between subject variability and population pharmacokinetics</subfield><subfield code="7">(dpeaa)DE-He213</subfield></datafield><datafield tag="700" ind1="1" ind2=" "><subfield code="a">Vajjah, Pavan</subfield><subfield code="e">verfasserin</subfield><subfield code="4">aut</subfield></datafield><datafield tag="700" ind1="1" ind2=" "><subfield code="a">Duffull, Stephen B.</subfield><subfield code="e">verfasserin</subfield><subfield code="4">aut</subfield></datafield><datafield tag="773" ind1="0" ind2="8"><subfield code="i">Enthalten in</subfield><subfield code="t">Journal of Pharmacokinetics and Biopharmaceutics</subfield><subfield code="d">Kluwer Academic Publishers-Plenum Publishers, 1973</subfield><subfield code="g">39(2012), 4 vom: 06. Juli, Seite 383-392</subfield><subfield code="w">(DE-627)SPR014694166</subfield><subfield code="7">nnns</subfield></datafield><datafield tag="773" ind1="1" ind2="8"><subfield code="g">volume:39</subfield><subfield code="g">year:2012</subfield><subfield code="g">number:4</subfield><subfield code="g">day:06</subfield><subfield code="g">month:07</subfield><subfield code="g">pages:383-392</subfield></datafield><datafield tag="856" ind1="4" ind2="0"><subfield code="u">https://dx.doi.org/10.1007/s10928-012-9256-2</subfield><subfield code="z">lizenzpflichtig</subfield><subfield code="3">Volltext</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_USEFLAG_A</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">SYSFLAG_A</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_SPRINGER</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">SSG-OLC-PHA</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_ILN_40</subfield></datafield><datafield tag="951" ind1=" " ind2=" "><subfield code="a">AR</subfield></datafield><datafield tag="952" ind1=" " ind2=" "><subfield code="d">39</subfield><subfield code="j">2012</subfield><subfield code="e">4</subfield><subfield code="b">06</subfield><subfield code="c">07</subfield><subfield code="h">383-392</subfield></datafield></record></collection>
|
| author |
Lagishetty, Chakradhar V. |
| spellingShingle |
Lagishetty, Chakradhar V. misc Non-nested and nested covariate models misc Predictable between subject variability misc Random between subject variability and population pharmacokinetics A reduction in between subject variability is not mandatory for selecting a new covariate |
| authorStr |
Lagishetty, Chakradhar V. |
| ppnlink_with_tag_str_mv |
@@773@@(DE-627)SPR014694166 |
| format |
electronic Article |
| delete_txt_mv |
keep |
| author_role |
aut aut aut |
| collection |
springer |
| remote_str |
true |
| illustrated |
Not Illustrated |
| topic_title |
A reduction in between subject variability is not mandatory for selecting a new covariate Non-nested and nested covariate models (dpeaa)DE-He213 Predictable between subject variability (dpeaa)DE-He213 Random between subject variability and population pharmacokinetics (dpeaa)DE-He213 |
| topic |
misc Non-nested and nested covariate models misc Predictable between subject variability misc Random between subject variability and population pharmacokinetics |
| topic_unstemmed |
misc Non-nested and nested covariate models misc Predictable between subject variability misc Random between subject variability and population pharmacokinetics |
| topic_browse |
misc Non-nested and nested covariate models misc Predictable between subject variability misc Random between subject variability and population pharmacokinetics |
| format_facet |
Elektronische Aufsätze Aufsätze Elektronische Ressource |
| format_main_str_mv |
Text Zeitschrift/Artikel |
| carriertype_str_mv |
cr |
| hierarchy_parent_title |
Journal of Pharmacokinetics and Biopharmaceutics |
| hierarchy_parent_id |
SPR014694166 |
| hierarchy_top_title |
Journal of Pharmacokinetics and Biopharmaceutics |
| isfreeaccess_txt |
false |
| familylinks_str_mv |
(DE-627)SPR014694166 |
| title |
A reduction in between subject variability is not mandatory for selecting a new covariate |
| ctrlnum |
(DE-627)SPR014710153 (SPR)s10928-012-9256-2-e |
| title_full |
A reduction in between subject variability is not mandatory for selecting a new covariate |
| author_sort |
Lagishetty, Chakradhar V. |
| journal |
Journal of Pharmacokinetics and Biopharmaceutics |
| journalStr |
Journal of Pharmacokinetics and Biopharmaceutics |
| lang_code |
eng |
| isOA_bool |
false |
| recordtype |
marc |
| publishDateSort |
2012 |
| contenttype_str_mv |
txt |
| container_start_page |
383 |
| author_browse |
Lagishetty, Chakradhar V. Vajjah, Pavan Duffull, Stephen B. |
| container_volume |
39 |
| format_se |
Elektronische Aufsätze |
| author-letter |
Lagishetty, Chakradhar V. |
| doi_str_mv |
10.1007/s10928-012-9256-2 |
| author2-role |
verfasserin |
| title_sort |
reduction in between subject variability is not mandatory for selecting a new covariate |
| title_auth |
A reduction in between subject variability is not mandatory for selecting a new covariate |
| abstract |
Abstract Population pharmacokinetic-pharmacodynamic analysis involves nonlinear hierarchical modelling where the mean response in a population and the variability in response from different sources are studied. It generally consists of two model hierarchies: a model for residual error and a model for heterogeneity termed between subject variance (BSV). The overall variability in a parameter within a population termed population parameter variance (PPV) consists of within subject variance (WSV) and BSV. Both these variances can further be split into random and predictable components. The predictable component of BSV (termed BSVP) is explained by covariates, individual characteristics e.g. weight. As BSVP increases, the remaining unpredictable (or random) between subject variability (BSVR) decreases since BSV = BSVP + BSVR, and BSV is a constant in any given data set. Since BSV and BSVR are estimated from the base and full covariate models, respectively, then BSVP = BSV−BSVR. The aim of this study was to explore the hypothesis, that a significant covariate may not always decrease BSVR. The specific aims were: (1) to explore circumstances where BSVR may not be reduced when adding a significantly correlated covariate and (2) to explore whether specific models for covariates may eliminate this anomaly when assessing BSVR. Simulations were performed using MATLAB (2011a) and estimation using NONMEM (ver 7.2) with FOCE and INTERACTION. A 1-compartment intravenous bolus PK model was used for simulation following a single unit dose (d = 1). The BSV of clearance [BSV(CL)] was described according to a log-normal distribution model with mean zero and variance %$ \omega^{ 2} %$. An additive random unexplained variability was assumed. Initially, we show through a simple simulation that BSVR can increase when a significantly correlated covariate is added to the model. We follow this with five simulation scenarios, A to E, that have various levels of correlation between the continuous covariate (Z) and CL ranging from 0 to 100 %. Each simulated scenario was replicated 100 times and estimated by a base model (i.e. without covariate addition) and six covariate models (M1–M6) which included non-nested (M1), nested (M2), and two types of interaction models for each of M1 and M2; non-nested interaction (M3, M5), nested interaction (M4, M6). Initially, through a motivating example we show that BSVR may not reduce even when there is 50 % correlation between the covariate Z and CL. It was found that with 0 % correlation M1, the non-nested covariate model (NNCM) resulted in negative BSVP (inflated BSVR) whereas M2, the nested covariate model (NCM), resulted in a calculated BSVP of zero. NNCM (M1) shows negative BSVP (BSVR > BSV) with correlation as high as 50 % and this model needs a minimum of 75 % correlation to show a positive BSVP. NCM (M2) shows positive but downwardly biased BSVP with 25, 50 and 75 % correlations. However, inclusion of a covariate–eta interaction term for both types of covariate models resulted in greater BSVP for 25, 50 and 75 % correlation scenarios compared to NNCM and NCM respectively. For 100 % correlation, it was found that covariate–eta interaction models show the same BSVP as the models without the interaction term, i.e. under perfect positive correlation all models perform similarly and correctly. It was found that a significantly correlated covariate may not reduce BSVR and in fact it may inflate the BSVR due to statistical misspecification of the covariate model. Incorporating statistical models that account for the covariate–eta interaction may be useful diagnostically in identifying the variability explained by covariates. |
| abstractGer |
Abstract Population pharmacokinetic-pharmacodynamic analysis involves nonlinear hierarchical modelling where the mean response in a population and the variability in response from different sources are studied. It generally consists of two model hierarchies: a model for residual error and a model for heterogeneity termed between subject variance (BSV). The overall variability in a parameter within a population termed population parameter variance (PPV) consists of within subject variance (WSV) and BSV. Both these variances can further be split into random and predictable components. The predictable component of BSV (termed BSVP) is explained by covariates, individual characteristics e.g. weight. As BSVP increases, the remaining unpredictable (or random) between subject variability (BSVR) decreases since BSV = BSVP + BSVR, and BSV is a constant in any given data set. Since BSV and BSVR are estimated from the base and full covariate models, respectively, then BSVP = BSV−BSVR. The aim of this study was to explore the hypothesis, that a significant covariate may not always decrease BSVR. The specific aims were: (1) to explore circumstances where BSVR may not be reduced when adding a significantly correlated covariate and (2) to explore whether specific models for covariates may eliminate this anomaly when assessing BSVR. Simulations were performed using MATLAB (2011a) and estimation using NONMEM (ver 7.2) with FOCE and INTERACTION. A 1-compartment intravenous bolus PK model was used for simulation following a single unit dose (d = 1). The BSV of clearance [BSV(CL)] was described according to a log-normal distribution model with mean zero and variance %$ \omega^{ 2} %$. An additive random unexplained variability was assumed. Initially, we show through a simple simulation that BSVR can increase when a significantly correlated covariate is added to the model. We follow this with five simulation scenarios, A to E, that have various levels of correlation between the continuous covariate (Z) and CL ranging from 0 to 100 %. Each simulated scenario was replicated 100 times and estimated by a base model (i.e. without covariate addition) and six covariate models (M1–M6) which included non-nested (M1), nested (M2), and two types of interaction models for each of M1 and M2; non-nested interaction (M3, M5), nested interaction (M4, M6). Initially, through a motivating example we show that BSVR may not reduce even when there is 50 % correlation between the covariate Z and CL. It was found that with 0 % correlation M1, the non-nested covariate model (NNCM) resulted in negative BSVP (inflated BSVR) whereas M2, the nested covariate model (NCM), resulted in a calculated BSVP of zero. NNCM (M1) shows negative BSVP (BSVR > BSV) with correlation as high as 50 % and this model needs a minimum of 75 % correlation to show a positive BSVP. NCM (M2) shows positive but downwardly biased BSVP with 25, 50 and 75 % correlations. However, inclusion of a covariate–eta interaction term for both types of covariate models resulted in greater BSVP for 25, 50 and 75 % correlation scenarios compared to NNCM and NCM respectively. For 100 % correlation, it was found that covariate–eta interaction models show the same BSVP as the models without the interaction term, i.e. under perfect positive correlation all models perform similarly and correctly. It was found that a significantly correlated covariate may not reduce BSVR and in fact it may inflate the BSVR due to statistical misspecification of the covariate model. Incorporating statistical models that account for the covariate–eta interaction may be useful diagnostically in identifying the variability explained by covariates. |
| abstract_unstemmed |
Abstract Population pharmacokinetic-pharmacodynamic analysis involves nonlinear hierarchical modelling where the mean response in a population and the variability in response from different sources are studied. It generally consists of two model hierarchies: a model for residual error and a model for heterogeneity termed between subject variance (BSV). The overall variability in a parameter within a population termed population parameter variance (PPV) consists of within subject variance (WSV) and BSV. Both these variances can further be split into random and predictable components. The predictable component of BSV (termed BSVP) is explained by covariates, individual characteristics e.g. weight. As BSVP increases, the remaining unpredictable (or random) between subject variability (BSVR) decreases since BSV = BSVP + BSVR, and BSV is a constant in any given data set. Since BSV and BSVR are estimated from the base and full covariate models, respectively, then BSVP = BSV−BSVR. The aim of this study was to explore the hypothesis, that a significant covariate may not always decrease BSVR. The specific aims were: (1) to explore circumstances where BSVR may not be reduced when adding a significantly correlated covariate and (2) to explore whether specific models for covariates may eliminate this anomaly when assessing BSVR. Simulations were performed using MATLAB (2011a) and estimation using NONMEM (ver 7.2) with FOCE and INTERACTION. A 1-compartment intravenous bolus PK model was used for simulation following a single unit dose (d = 1). The BSV of clearance [BSV(CL)] was described according to a log-normal distribution model with mean zero and variance %$ \omega^{ 2} %$. An additive random unexplained variability was assumed. Initially, we show through a simple simulation that BSVR can increase when a significantly correlated covariate is added to the model. We follow this with five simulation scenarios, A to E, that have various levels of correlation between the continuous covariate (Z) and CL ranging from 0 to 100 %. Each simulated scenario was replicated 100 times and estimated by a base model (i.e. without covariate addition) and six covariate models (M1–M6) which included non-nested (M1), nested (M2), and two types of interaction models for each of M1 and M2; non-nested interaction (M3, M5), nested interaction (M4, M6). Initially, through a motivating example we show that BSVR may not reduce even when there is 50 % correlation between the covariate Z and CL. It was found that with 0 % correlation M1, the non-nested covariate model (NNCM) resulted in negative BSVP (inflated BSVR) whereas M2, the nested covariate model (NCM), resulted in a calculated BSVP of zero. NNCM (M1) shows negative BSVP (BSVR > BSV) with correlation as high as 50 % and this model needs a minimum of 75 % correlation to show a positive BSVP. NCM (M2) shows positive but downwardly biased BSVP with 25, 50 and 75 % correlations. However, inclusion of a covariate–eta interaction term for both types of covariate models resulted in greater BSVP for 25, 50 and 75 % correlation scenarios compared to NNCM and NCM respectively. For 100 % correlation, it was found that covariate–eta interaction models show the same BSVP as the models without the interaction term, i.e. under perfect positive correlation all models perform similarly and correctly. It was found that a significantly correlated covariate may not reduce BSVR and in fact it may inflate the BSVR due to statistical misspecification of the covariate model. Incorporating statistical models that account for the covariate–eta interaction may be useful diagnostically in identifying the variability explained by covariates. |
| collection_details |
GBV_USEFLAG_A SYSFLAG_A GBV_SPRINGER SSG-OLC-PHA GBV_ILN_40 |
| container_issue |
4 |
| title_short |
A reduction in between subject variability is not mandatory for selecting a new covariate |
| url |
https://dx.doi.org/10.1007/s10928-012-9256-2 |
| remote_bool |
true |
| author2 |
Vajjah, Pavan Duffull, Stephen B. |
| author2Str |
Vajjah, Pavan Duffull, Stephen B. |
| ppnlink |
SPR014694166 |
| mediatype_str_mv |
c |
| isOA_txt |
false |
| hochschulschrift_bool |
false |
| doi_str |
10.1007/s10928-012-9256-2 |
| up_date |
2024-07-04T02:46:54.612Z |
| _version_ |
1803614908174041088 |
| fullrecord_marcxml |
<?xml version="1.0" encoding="UTF-8"?><collection xmlns="http://www.loc.gov/MARC21/slim"><record><leader>01000caa a22002652 4500</leader><controlfield tag="001">SPR014710153</controlfield><controlfield tag="003">DE-627</controlfield><controlfield tag="005">20230519141606.0</controlfield><controlfield tag="007">cr uuu---uuuuu</controlfield><controlfield tag="008">201006s2012 xx |||||o 00| ||eng c</controlfield><datafield tag="024" ind1="7" ind2=" "><subfield code="a">10.1007/s10928-012-9256-2</subfield><subfield code="2">doi</subfield></datafield><datafield tag="035" ind1=" " ind2=" "><subfield code="a">(DE-627)SPR014710153</subfield></datafield><datafield tag="035" ind1=" " ind2=" "><subfield code="a">(SPR)s10928-012-9256-2-e</subfield></datafield><datafield tag="040" ind1=" " ind2=" "><subfield code="a">DE-627</subfield><subfield code="b">ger</subfield><subfield code="c">DE-627</subfield><subfield code="e">rakwb</subfield></datafield><datafield tag="041" ind1=" " ind2=" "><subfield code="a">eng</subfield></datafield><datafield tag="100" ind1="1" ind2=" "><subfield code="a">Lagishetty, Chakradhar V.</subfield><subfield code="e">verfasserin</subfield><subfield code="4">aut</subfield></datafield><datafield tag="245" ind1="1" ind2="2"><subfield code="a">A reduction in between subject variability is not mandatory for selecting a new covariate</subfield></datafield><datafield tag="264" ind1=" " ind2="1"><subfield code="c">2012</subfield></datafield><datafield tag="336" ind1=" " ind2=" "><subfield code="a">Text</subfield><subfield code="b">txt</subfield><subfield code="2">rdacontent</subfield></datafield><datafield tag="337" ind1=" " ind2=" "><subfield code="a">Computermedien</subfield><subfield code="b">c</subfield><subfield code="2">rdamedia</subfield></datafield><datafield tag="338" ind1=" " ind2=" "><subfield code="a">Online-Ressource</subfield><subfield code="b">cr</subfield><subfield code="2">rdacarrier</subfield></datafield><datafield tag="520" ind1=" " ind2=" "><subfield code="a">Abstract Population pharmacokinetic-pharmacodynamic analysis involves nonlinear hierarchical modelling where the mean response in a population and the variability in response from different sources are studied. It generally consists of two model hierarchies: a model for residual error and a model for heterogeneity termed between subject variance (BSV). The overall variability in a parameter within a population termed population parameter variance (PPV) consists of within subject variance (WSV) and BSV. Both these variances can further be split into random and predictable components. The predictable component of BSV (termed BSVP) is explained by covariates, individual characteristics e.g. weight. As BSVP increases, the remaining unpredictable (or random) between subject variability (BSVR) decreases since BSV = BSVP + BSVR, and BSV is a constant in any given data set. Since BSV and BSVR are estimated from the base and full covariate models, respectively, then BSVP = BSV−BSVR. The aim of this study was to explore the hypothesis, that a significant covariate may not always decrease BSVR. The specific aims were: (1) to explore circumstances where BSVR may not be reduced when adding a significantly correlated covariate and (2) to explore whether specific models for covariates may eliminate this anomaly when assessing BSVR. Simulations were performed using MATLAB (2011a) and estimation using NONMEM (ver 7.2) with FOCE and INTERACTION. A 1-compartment intravenous bolus PK model was used for simulation following a single unit dose (d = 1). The BSV of clearance [BSV(CL)] was described according to a log-normal distribution model with mean zero and variance %$ \omega^{ 2} %$. An additive random unexplained variability was assumed. Initially, we show through a simple simulation that BSVR can increase when a significantly correlated covariate is added to the model. We follow this with five simulation scenarios, A to E, that have various levels of correlation between the continuous covariate (Z) and CL ranging from 0 to 100 %. Each simulated scenario was replicated 100 times and estimated by a base model (i.e. without covariate addition) and six covariate models (M1–M6) which included non-nested (M1), nested (M2), and two types of interaction models for each of M1 and M2; non-nested interaction (M3, M5), nested interaction (M4, M6). Initially, through a motivating example we show that BSVR may not reduce even when there is 50 % correlation between the covariate Z and CL. It was found that with 0 % correlation M1, the non-nested covariate model (NNCM) resulted in negative BSVP (inflated BSVR) whereas M2, the nested covariate model (NCM), resulted in a calculated BSVP of zero. NNCM (M1) shows negative BSVP (BSVR > BSV) with correlation as high as 50 % and this model needs a minimum of 75 % correlation to show a positive BSVP. NCM (M2) shows positive but downwardly biased BSVP with 25, 50 and 75 % correlations. However, inclusion of a covariate–eta interaction term for both types of covariate models resulted in greater BSVP for 25, 50 and 75 % correlation scenarios compared to NNCM and NCM respectively. For 100 % correlation, it was found that covariate–eta interaction models show the same BSVP as the models without the interaction term, i.e. under perfect positive correlation all models perform similarly and correctly. It was found that a significantly correlated covariate may not reduce BSVR and in fact it may inflate the BSVR due to statistical misspecification of the covariate model. Incorporating statistical models that account for the covariate–eta interaction may be useful diagnostically in identifying the variability explained by covariates.</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Non-nested and nested covariate models</subfield><subfield code="7">(dpeaa)DE-He213</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Predictable between subject variability</subfield><subfield code="7">(dpeaa)DE-He213</subfield></datafield><datafield tag="650" ind1=" " ind2="4"><subfield code="a">Random between subject variability and population pharmacokinetics</subfield><subfield code="7">(dpeaa)DE-He213</subfield></datafield><datafield tag="700" ind1="1" ind2=" "><subfield code="a">Vajjah, Pavan</subfield><subfield code="e">verfasserin</subfield><subfield code="4">aut</subfield></datafield><datafield tag="700" ind1="1" ind2=" "><subfield code="a">Duffull, Stephen B.</subfield><subfield code="e">verfasserin</subfield><subfield code="4">aut</subfield></datafield><datafield tag="773" ind1="0" ind2="8"><subfield code="i">Enthalten in</subfield><subfield code="t">Journal of Pharmacokinetics and Biopharmaceutics</subfield><subfield code="d">Kluwer Academic Publishers-Plenum Publishers, 1973</subfield><subfield code="g">39(2012), 4 vom: 06. Juli, Seite 383-392</subfield><subfield code="w">(DE-627)SPR014694166</subfield><subfield code="7">nnns</subfield></datafield><datafield tag="773" ind1="1" ind2="8"><subfield code="g">volume:39</subfield><subfield code="g">year:2012</subfield><subfield code="g">number:4</subfield><subfield code="g">day:06</subfield><subfield code="g">month:07</subfield><subfield code="g">pages:383-392</subfield></datafield><datafield tag="856" ind1="4" ind2="0"><subfield code="u">https://dx.doi.org/10.1007/s10928-012-9256-2</subfield><subfield code="z">lizenzpflichtig</subfield><subfield code="3">Volltext</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_USEFLAG_A</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">SYSFLAG_A</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_SPRINGER</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">SSG-OLC-PHA</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_ILN_40</subfield></datafield><datafield tag="951" ind1=" " ind2=" "><subfield code="a">AR</subfield></datafield><datafield tag="952" ind1=" " ind2=" "><subfield code="d">39</subfield><subfield code="j">2012</subfield><subfield code="e">4</subfield><subfield code="b">06</subfield><subfield code="c">07</subfield><subfield code="h">383-392</subfield></datafield></record></collection>
|
| score |
7.400177 |