BJA Education, 16 (3): 92–97 (2016)

doi: 10.1093/bjaceaccp/mkv021
Advance Access Publication Date: 8 June 2015

Matrix reference
1A02, 3I00

Principles of total intravenous anaesthesia: basic

pharmacokinetics and model descriptions

Downloaded from http://bjaed.oxfordjournals.org/ by Syed Zeeshan Javaid Hashmi on December 17, 2016
Z Al-Rifai MBChB (Hons) MPharm (Hons) FRCA1 and D Mulvey BSc
(Hons) MBBS MD FRCA2, *
1
ST4 Dual Speciality Trainee in Anaesthesia and Critical Care, Department of Anaesthesia, Royal Derby Hospital,
Uttoxeter Road, Derby DE22 3NN, UK, and 2Consultant Anaesthetist, Department of Anaesthesia, Royal Derby
Hospital, Uttoxeter Road, Derby DE22 3NN, UK
*To whom correspondence should be addressed. E-mail: david.mulvey@nhs.net

Any combination of hypnotics (with or without analgesics)

Key points can be used to achieve a desired clinical endpoint. This hetero-
geneity confounds the interpretation of TIVA outcome data as
• Competency in total i.v. anaesthesia (TIVA) allows
no technique defines ‘ideal’ TIVA. Suboptimal techniques are
safe management of general anaesthesia in
often included in outcome analyses, but the results are general-
patients with malignant hyperthermia risk.
ised to all methodologies.1–3 Poor understanding of the pharma-
• Poor understanding of the pharmacokinetics of cokinetics underlying TIVA has caused accidental awareness as
target-controlled infusion (TCI)/TIVA practice has documented in the Fifth National Audit Project on accidental
contributed to accidental anaesthetic awareness awareness during general anaesthesia (NAP5) report.3
as reported by NAP5.
• There is no defined ‘ideal’ TIVA technique, but
Choice of agents
co-administration of propofol and remifentanil by
TCIs approaches this goal. Drugs with fast onset and offset times are most useful for balancing
adequate hypnosis/analgesia with rapid recovery. The decline in
• A variety of pharmacokinetic models for propofol
the plasma concentration of most i.v. agents slows as the duration
and remifentanil have been described, but only a
of infusion increases (‘context-sensitive half-time’—CSHT) and im-
few have been implemented in commercial infu-
pairs recovery. Propofol and remifentanil demonstrate short or
sion devices.
minimal CSHT unlike other i.v. agents. For this article, ‘ideal’ TIVA
• All models incorporate assumptions and elements constitutes the co-administration of these agents by target-con-
of inaccuracy in the prediction of plasma and trolled infusion (TCI). This approach exploits their known synergy
effect-site targets. However, inter-individual vari- in obtunding responses to noxious stimuli.4
ability in pharmacodynamic response represents a
more challenging aspect of using TIVA.
The three-compartment model
After an i.v. bolus, the plasma concentration of a typical drug fol-
Total i.v. anaesthesia (TIVA) describes the maintenance of gen- lows an exponential decline in three distinct phases (Fig. 1).
eral anaesthesia without inhaled hypnotics. Some indications These observations are explained by distribution of drug between
for TIVA are given in Table 1. Competency in TIVA is vital for a central compartment (V1, principally plasma) and two com-
safe management of patients with malignant hyperthermia partments which equilibrate rapidly (V2, well-perfused tissue
risk who require general anaesthesia. like muscle) and slowly (V3, mainly fatty tissue) (Fig. 2).

© The Author 2015. Published by Oxford University Press on behalf of the British Journal of Anaesthesia. All rights reserved.
For Permissions, please email: journals.permissions@oup.com

92
 Principles of TIVA

Table 1 Indications for using TIVA in pharmacokinetic pumps use more exacting analytical solu-
tions). Once compartment V1 is filled by the bolus, the subse-
quent infusion rate compensates for rapid and slow transfer of
• Malignant hyperthermia risk drug to V2 and V3, and drug elimination from V1 as described by
• Long QT Syndrome (QTc≥500 ms) the rate constant K10 (rate constant for drug elimination from the
• History of severe PONV central compartment in a pharmacokinetic model). When the
• ‘Tubeless’ ENT and thoracic surgery three compartments reach steady-state concentration (>20 h
• Patients with anticipated difficult intubation/extubation for propofol), the infusion rate slows to match elimination only.
• Neurosurgery—to limit intracranial volume Without an appropriate bolus, a constant propofol infusion at
• Surgery requiring neurophysiological monitoring 10 mg kg−1 h−1 requires 40–90 min (dependent upon which ki-
• Myasthenia gravis/neuromuscular disorders to avoid NMBs netic model is used for calculation) to achieve a clinically useful
• Anaesthesia in non-theatre environments plasma concentration of 4 μg ml−1 in an 85 kg adult male. It is like-
• Transfer of anaesthetised patient between environments ly that an inadequate clinical effect would be observed in the in-
• Day-case surgery terim as was reported by NAP5.3
• Trainee teaching
• Patient choice

Downloaded from http://bjaed.oxfordjournals.org/ by Syed Zeeshan Javaid Hashmi on December 17, 2016
The TCI system
The key components are:

• User interface
• Microprocessor(s) with pharmacokinetic software
• Infusion pump which delivers up to 1200 ml h−1
• Visual and audible safety systems and alarms

A typical system calculates the bolus dose and speed of subse-

quent infusion required to maintain the targeted plasma drug
concentration (Cpt). Calculations are repeated every 10 s and
the infusion rate adjusted until Cpt is achieved. Diffusion of
drug from plasma to brain occurs exponentially with a first-
order rate constant (ke0, see below). The half-time (T1/2) for the
process is calculated as T1/2 = [ln (2)/ke0] and equilibration occurs
after 4–5 half-times.8 If Cpt is subsequently increased, an add-
itional bolus is given to fill V1 and the infusion rate increases
to match additional transfer and elimination at the higher
concentration.
When Cpt is decreased, the infusion stops until the plasma
Fig 1 Plasma concentration vs time curve demonstrating tri-exponential decline
after a bolus injection. In conventional pharmacokinetic terminology, these are
concentration declines to the new target and is restarted at a
phases A–C. The plasma concentration at time t(Ct) may be derived from lower rate. Diffusion of drug from the brain occurs with the
CðtÞ ¼ Aeαt þ Beβt þ Ceγt , where t is time since i.v. bolus. C, concentration same half-time.
after a bolus dose. A–C represent the phase coefficients which sum to the
plasma concentration after an i.v. bolus. α, β, and γ represent phase rate
constants. e, natural logarithm. Common TCI models
The differences in published models for propofol and remifenta-
Mathematical analysis allows the compartment volumes and nil result from methodological aspects and limitations of the ori-
rate constants for drug transfer between them to be calculated.5 ginal studies. Relatively few of these solutions are implemented
Detailed discussion of the modelling process has been published in commercial infusion devices (Table 2). Currently, there is no
in this journal.6,7 evidence to support the use of one model in preference to an-
other and all have proved reliable in clinical practice. All have
similar performance in terms of the accuracy and stability of pre-
Target-controlled infusions
dicted plasma (Cp) and effect-site (Ce) concentrations.
Adequacy of TIVA depends on the maintenance of brain propofol
and remifentanil concentrations which are clinically appropriate
Propofol
and in equilibrium with levels in the plasma. The best way to
achieve this state is by TCI from dedicated pharmacokinetic The key pharmacokinetic parameters for the adult Marsh and
pumps. These devices solve the complex equations which de- Schnider plasma-targeting models are shown in Table 2. The
scribe the distribution of agents between compartments and major difference is the volume of V1 (Marsh 19.4 litre vs Schnider
allow for rapid adjustments in targets to achieve the desired clin- 4.27 litre for an 85 kg individual), and therefore a bolus adminis-
ical effect. Manual infusion regimes are prone to errors in the cal- tered as mg kg−1 causes a four-fold difference in calculated peak
culation and implementation of the required changes in infusion plasma concentrations (Fig. 3).
rate as reported by NAP5.3 The Marsh model ignores age and scales the volumes of V1−3
linearly to patient weight. An identical bolus dose is adminis-
tered to all patients of a given body mass for any chosen Cpt.
Principles of TCI This delivery contrasts with non-TIVA practice where the anaes-
A bolus/elimination/transfer (BET) principle is used to approxi- thetist usually adjusts dosage for patient age and likely pharma-
mate a constant plasma level of drug (however, the algorithms codynamic response. Age is input to the TCI pump only to ensure

BJA Education | Volume 16, Number 3, 2016 93

Principles of TIVA

Downloaded from http://bjaed.oxfordjournals.org/ by Syed Zeeshan Javaid Hashmi on December 17, 2016
Fig 2 Three-compartment model showing the various compartments and their associated rate constants. V1−3 represents the compartment volumes. K12 represents the
rate constant between V1 and V2, K21 between V2 and V1, etc. K10 represents the rate constant for drug elimination from the central compartment. ke0 is the rate constant
for equilibration between plasma and effect-site concentrations.

Table 2 Comparison of the pharmacokinetic parameters for the main clinical effect. Consequently, the Schnider model can only be re-
TCI models implemented in commercial infusion devices. V1, central commended for use in effect-site targeting mode (see below) as
compartment; V2, rapidly equilibrating compartment; V3, slowly larger bolus doses are utilised.
equilibrating compartment; K10, rate constant for drug elimination Small differences between the Paedfusor and Kataria paediat-
from the central compartment in a pharmacokinetic model; Kxy and
ric models are shown in Table 2. Both use weight as the key pa-
Kyx, rate constants for drug transfer from compartment x to
compartment y or the reverse direction; LBM, lean body mass
tient characteristic for scaling the volumes of V1−3. The Kataria
model is validated for use in patients aged 3–16 yr with a min-
Model Fixed Variable Parameter imum weight of 15 kg. The Paedfusor model is a variant of the
parameters parameters determined by Marsh kinetics for patients 1–16 yr of age, and also uses weight
to calculate the elimination constant K10. It features non-linear
Marsh All rate V1,2,3 Weight scaling of V1 volume as age exceeds 12 yr. Extrapolation of
constants these models to patients outside of the described patient charac-
Schnider V1 = 4.7 litre V2 Age
teristics is not recommended due to increased pharmacokinetic
V3, K13, K31 K12, K21 Age
differences, but is commonly practised.
K10 Age, weight,
A recently described allometric scaling model promises im-
LBM
proved utility in propofol administration.9 Allometry relates bio-
Paedfusor All rate V1,2,3 Weight
logical activity to proportional rather than absolute changes in
constants K10 Weight
except K10
body size. In the new model, 10 927 blood propofol concentra-
Kataria All rate V1,2,3 Weight tions were aggregated from 660 subjects (500 patients, 160
constants healthy volunteers) recruited to 21 separate studies. This con-
Minto V3 = 5.42 litre V1 and V2 and Age, LBM trasts with the limited data used to generate the Marsh and
rate constants Schnider kinetics. Body composition in the new model is re-
flected in the calculation of the volumes of V2 and V3 which in
turn are the primary determinants of their associated rate con-
stants, not absolute body mass. This complex analysis allows
the model to be used for patients aged 3 months to 88 yr and
that the patient is ≥16 yr and that the use of this model is appro- weighing 5–160 kg, but needs further clinical validation before
priate. For less robust patients, it is better to start the pump at a it becomes commercially available.
lower Cpt and increase the target incrementally until a desired
clinical effect is obtained.
Remifentanil
Although the Schnider model adjusts some of the pharmaco-
kinetic parameters for age (Table 2), this does not necessarily The Minto model for remifentanil is popular because it is applic-
constrain the patient’s pharmacodynamic response. A sex- able to a wide range of patient characteristics. Age is used
specific lean body mass (LBM) is calculated and used to adjust for calculation of pharmacokinetic parameters (Table 2) but in
the elimination rate constant K10. Because a small fixed volume common with the Schnider model, this adjustment does not
for V1 is used, lower doses of propofol are required to achieve a influence pharmacodynamic response. A sex-specific LBM is
given Cpt compared with Marsh (Table 3). In many instances, calculated and used to fine tune some of the parameters to the
this bolus is inappropriately small and results in an inadequate patient.

94 BJA Education | Volume 16, Number 3, 2016

 Principles of TIVA

Downloaded from http://bjaed.oxfordjournals.org/ by Syed Zeeshan Javaid Hashmi on December 17, 2016
Fig 3 Comparison of propofol concentrations predicted for a 40-yr-old male, 178 cm tall, and weighing 85 kg by the Marsh (upper panels) and Schnider (lower panels)
models using Tivatrainer 9 software (www.eurosiva.eu). Left-hand panels show plasma targeting (Cpt) mode; right-hand panels show effect-site targeting (Cet). The
target was set at 4 μg ml−1 and all diagrams have the same time scale in minutes on the x-axis. Note the major difference in plasma concentration predicted in Cet
modes—this is largely due to the difference in the volume of the central compartment V1 assumed by each model. Differences in the time to equilibration of the
plasma and effect-site concentrations for the Marsh model are due to the different ke0 utilised in each calculation. The ke0s used in the effect-site models are based
on a TTPE of 1.6 min after the bolus dose. Red line, plasma concentration; orange line, chosen target level; green line, effect-site concentration; vertical white lines
represent the infusion rate of TCI pump.

Effect-site targeting drug transfer into the brain. The bolus dose required to produce
this plasma overshoot is governed by the volume of V1 used in
Propofol
the calculation and hence the model chosen. Once the bolus is
Clinicians regularly but unintentionally use effect-site targeting given, the TCI pump stops infusing until equilibration occurs at
in non-TIVA practice to rapidly achieve unconsciousness. The the relevant TTPE and then re-starts at a rate matching inter-
high plasma concentration generated by a large bolus of propofol compartmental transfer and elimination (also right-hand panels
causes fast diffusion of drug into the brain and rapid onset of in Fig. 3).
the desired effect. However, propofol diffusion continues until However, brain propofol concentrations cannot be measured
the brain concentration has equilibrated with the diminishing in vivo and the rate constant required for direct calculation of
plasma level (Fig. 3, right-hand panels). This occurs 1.6–3.9 min plasma overshoot is unknowable. Instead, drug concentration
after injection irrespective of the size of the bolus dose9 (the in a theoretical surrogate called the effect-site is used for math-
‘time to peak effect’, TTPE, as predicted by the Schnider and ematical analysis of effect and prediction of plasma overshoot.10
Marsh models, respectively). Depending on the patient’s idiosyn- This virtual compartment is assumed to have negligible volume
cratic sensitivity to propofol, this peak brain level can cause compared with V1 and causes little perturbation in plasma con-
unwanted cardiovascular instability (reflecting an effect-site centration at equilibrium. A rate constant called ke0 (Fig. 2) de-
overshoot). Effect-site targeting achieves unconsciousness rapidly scribes equilibration of effect-site concentration with plasma
without effect-site overshoot provided that the target effect-site propofol levels but has to be derived indirectly in experimental
drug concentration (Cet) is appropriate for the patient’s physical studies.
status. A measure of dynamic anaesthetic effect, typically a pro-
A peak blood concentration (called the plasma overshoot) cessed EEG signal, is recorded simultaneously with measured
generates the desired Cet at equilibrium (Fig. 3, right-hand pa- plasma propofol concentrations in volunteer subjects. A numeric
nels) and can in theory be calculated from the rate constant for value for ke0 can then be derived to match the timing of the

BJA Education | Volume 16, Number 3, 2016 95

Principles of TIVA

Table 3 Comparison of the dose of propofol administered to a 40-yr-old male, 178 cm tall, and 85 kg in weight when a 4 μg ml−1 target
concentration is set. The table shows the initial bolus dose, the time to equilibration of plasma and effect-site concentrations, and the cumulative
doses at two time points. Time to equilibration is equivalent to TTPE with effect-site targeting. Data derived from Tivatrainer 9 software (www.
eurosiva.eu). Greyed areas show non-applicable data

Comparison of adult propofol models

Model Plasma targeting Cpt Effect-site targeting Cet

Initial Time to Cumulative dose Initial Time to Cumulative dose

bolus equilibrium bolus equilibrium
At After 30 min of At After 30 min of
(mg) (min) (mg) (min)
equilibrium infusion (mg) equilibrium infusion (mg)
(mg) (mg)

Marsh 86 16 360 570

‘Modified 119 1.6 119 574
Marsh’
Schnider 18 9.8 142 414 73 1.6 73 444

Downloaded from http://bjaed.oxfordjournals.org/ by Syed Zeeshan Javaid Hashmi on December 17, 2016
calculated peak Ce to the observed maximum EEG effect. This ke0 (Table 3) where unwanted effect overshoot is common. The
is linked to rate of decay of drug concentration in V1 and hence to propofol TCI models described above prove highly effective in
the specific pharmacokinetic model used to administer the combination with remifentanil TCI.
agent. Once established, ke0 allows manipulation of plasma
overshoot to achieve a particular Cet.
For any given pharmacokinetic model, a large numeric value Remifentanil
of ke0 will predict a more rapid increase in Ce and allow a smaller The Minto model for remifentanil can be used in the Cet mode,
initial bolus dose to be given. Similarly, the decline in Ce to a level but the ke0 used is derived from studies of EEG parameters as a
representing recovery from anaesthesia will be predicted to occur measure of effect. It must be remembered that an EEG parameter
more quickly. does not necessarily equate with the onset of analgesic action.
Effect-site targeting is the only approach recommended for The increased plasma remifentanil concentrations required
the Schnider model as larger bolus doses are required to achieve in the Cet mode can be associated with a higher likelihood of
plasma overshoot (Table 3). Because this model incorporates chest wall rigidity and severe bradycardia via non-vagal mechan-
more patient characteristics, it has been recommended for use isms, so an incremental approach to remifentanil Cet may be a
in the elderly and less robust patient. However, it is always better reasonable technique.
to start at a low Cet in frail individuals and increase the target in-
crementally until the desired response is obtained.
The original description of the Schnider Cet model matched Conclusion
peak kinetic and dynamic parameters at a TTPE of 1.6 min and
allowed for small variations in ke0 between individuals based Currently, there is no ‘best’ TCI model for propofol. The clinician
on their idiosyncratic pharmacokinetics.10 It is also possible to should become familiar with the model which matches the
use the ‘average’ ke0 published in this study as a fixed parameter patient characteristics of their usual patient population. All phar-
and allow small variations in TTPE between patients.11 Pump macokinetic models have inherent assumptions which generate
manufacturers implement one particular approach in their elements of inaccuracy in prediction. However, inter-individual
equipment. Consequently, a clinician using different pump variability in pharmacodynamic response represents a more
brands may observe some differences in the bolus dose adminis- challenging aspect of using TIVA. Close clinical monitoring of
tered for a desired Cet in similar individuals.12 Care must be taken the patient remains an important part of the anaesthetist’s role.
to ensure that the expected clinical response is actually obtained.
The original ‘Diprifusor’ Marsh Cpt model did not have a ke0
and could not be used in the Cet mode. Subsequently, a ke0 was Educational video
assigned to allow calculation of Ce for information only, and The European Society for Intravenous Anaesthesia (EuroSIVA)
the selected value predicts a TTPE of 3.9–4.5 min after a typical provides an educational video on the basic pharmacokinetics of
2–3 mg kg−1 bolus. This timing may seem lengthy, but the predic- a simple infusion at the following link: https://www.youtube.
tion of Ce by this model has been validated in a patient study.13 com/watch?v=6U_K-ToHRvs (accessed 1 May 2015).
Attempts have been made to find the ‘best’ ke0 for enabling the
Marsh model in the Cet mode, and a variety of solutions have
been published. A method using a TTPE of 1.6 min to derive ke0 Acknowledgement
has been implemented in some commercial TCI devices as the
The authors would like to thank Dr Frank Engbers for his helpful
‘Modified Marsh’ model.14
comments on this manuscript.
For both Marsh and Schnider models, larger bolus doses of
propofol are used in the Cet mode at any given numeric target
compared with the same model in the Cpt mode. Marsh always
administers a larger bolus dose than Schnider in either mode
Declaration of interest
principally due to the significant difference in the volume of V1 D.M. is a member of the Committee of SIVA, the UK Society for
used in calculations. However, the absolute mass of agent may Intravenous Anaesthesia (www.siva.ac.uk) (accessed 1 May
seem small compared with that used in non-TIVA practice 2015).

96 BJA Education | Volume 16, Number 3, 2016

 Principles of TIVA

MCQs 6. Hill SA. Pharmacokinetics of drug infusions. Contin Educ

Anaesth Crit Care Pain 2004; 4: 76–80
The associated MCQs (to support CME/CPD activity) can be
7. Roberts F, Freshwater-Turner D. Pharmacokinetics and
accessed at https://access.oxfordjournals.org by subscribers to
anaesthesia. Contin Educ Anaesth Crit Care Pain 2007; 7:
BJA Education.
25–9
8. Cortınez LI. What is the ke0 and what does it tell me about
propofol? Anaesthesia 2014; 69: 399–402
References
9. Eleveld DJ, Proost JH, Cortínez LI et al. A general purpose phar-
1. Errando CL, Sigl JC, Robles M et al. Awareness with recall dur- macokinetic model for propofol. Anesth Analg 2014; 118:
ing general anaesthesia: a prospective observational evalu- 1221–37
ation of 4001 patients. Br J Anaesth 2008; 101: 178–85 10. Minto CF, Schnider TW, Gregg KM et al. Using the time of
2. Zhang C, Xu L, Ma YQ et al. Bispectral index monitoring maximum effect-site concentration to combine pharmaco-
prevents awareness during total intravenous anesthesia: a kinetics and pharmacodynamics. Anesthesiology 2003; 99:
prospective, randomized, double-blinded, multi-center con- 324–33
trolled trial. Chin Med J 2011; 124: 3664–9 11. Absalom A, Struys MM. Overview of Target Controlled Infusions
3. Nimmo AF, Cook TM. Total intravenous anaesthesia. 5th and Total Intravenous Anaesthesia, 2nd Edn. Gent: Academia

Downloaded from http://bjaed.oxfordjournals.org/ by Syed Zeeshan Javaid Hashmi on December 17, 2016
National Audit Project (NAP5). Accidental Awareness during Press, 2007.
General Anaesthesia in the United Kingdom and Ireland Report 12. Absalom AR, Mani V, De Smet T, Struys MM. Pharmacokinetic
and Findings—Chapter 18. The Royal College of Anaesthetists models for propofol—defining and illuminating the devil in
and the Association of Anaesthetists of Great Britain and the detail. Br J Anaesth 2009; 103: 26–37
Ireland, 2014. 13. Barakat AR, Sutcliffe N, Schwab M. Effect-site concentration
4. Vuyk J, Lim T, Engbers FH et al. The pharmacodynamic inter- during propofol TCI sedation: a comparison of sedation
action of propofol and alfentanil during lower abdominal score with two pharmacokinetic models. Anaesthesia 2007;
surgery in women. Anesthesiology 1995; 83: 8–22 62: 661–6
5. Engbers FH. Basic pharmacokinetics. In: Vuyk J, Schraag S, 14. Struys MM, De Smet T, Depoorter B et al. Comparison of
eds. Advances in Modelling and Clinical Applications of plasma compartment versus two methods for effect com-
Intravenous Anaesthesia. New York: Kluwer Academic/Plenum partment-controlled target-controlled infusion for propofol.
Publishers, 2003; 3–18 Anesthesiology 2000; 92: 399–406

BJA Education | Volume 16, Number 3, 2016 97