This was superseded first by electronic drip counters, which were subsequently made to control the adjustable clamp on a gravity-fed giving set (infusion controllers) and, ultimately, by microprocessor-controlled infusion pumps (Fig. 19.1), able to generate flow irrespective of the effects of gravity, and incorporating many features such as sensing infusion line pressure and the facility for being programmed in a variety of units and even languages to give stepped infusions based on patient weight. The first devices to not rely on a drip counter used a syringe type cassette to reliably control the volumes delivered and were hence known as ‘volumetric pumps’ to differentiate them (Fig. 19.2). This design is less common now, and most devices rely on a peristaltic mechanism to push fluid along. The term should now probably be made obsolete to be replaced by ‘infusion pump’.

Figure 19.1 Evolution of infusion pumps, (MP = microprocessor).

Figure 19.2 Principle of a cassette type (‘volumetric’) infusion pump.

Syringe drivers (also called syringe pumps) have traditionally been used for more accurate control of smaller volumes of infusion. The very first were simple clockwork devices designed to drive the plunger of a syringe at an even rate usually over 24 h. Syringe pumps now use the same technology as other infusion pumps and all these devices are considered together here.

1996 saw further progress with the commercial introduction of the Diprifusor system. This, a proprietary ‘chip’ added into a microprocessor-controlled syringe driver, allows the delivery of specially packaged Diprivan (propofol) by target-controlled infusion (TCI), and is discussed in greater detail below and also again later in the chapter. Licensing issues initially hampered the arrival of ‘open label TCI’ (see below), but at least four manufacturers now produce devices with open label pharmacokinetic algorithms capable of infusing at least the drugs propofol and remifentanil by TCI. Although it would be logical to use such systems for any drug (anaesthetic or otherwise) with a short half-life that needs to be given by continuous infusion, the development costs and licensing concerns for this route of administration for any given drug mean that it is only rarely commercially viable.

Microprocessor controlled/software driven

Modern electrically powered infusion devices use a stepper motor (see below). This is controlled by a microprocessor which ultimately simply varies the flow rate of the device from between 0 to 999 ml/h (infusion pump) or 1200 ml/h (for a syringe driver, depending on syringe size) at any given time.

Depending on the intended purpose of the infusion device the processor can be made to accept commands from the user for controlling the infusion in many different ways:

• simple ml/h request giving rise to the simplest infusion pump, say for ward use

• unit of drug per unit patient weight per unit time with variable units used for each; for example mg/kg/h or µg/kg/min. Clearly when dosing drug in this manner, the drug concentration in use and the patient weight must be input to the pump to allow automatic calculation of the flow rate. Such pumps may be used in the intensive care unit setting, but are more often termed ‘anaesthesia pumps’ as liberal discretion for choice of units is now usually only entrusted to anaesthetists

• simple infusion rate with an additional bolus volume when commanded by a separately attached control handle, thus constituting a PCA pump

• the in-built processor (or an additional piggybacked processor) can contain algorithms for the pharmacokinetics of particular drugs, allowing automated drug delivery based on achieving any desired theoretical patient plasma or effect site concentration of that drug. Such a configuration is termed a ‘target-controlled infusion (TCI) pump’. These systems are currently only commercially available for use in anaesthesia and are principally limited to four drugs: propofol, remifentanil, sufentanil and alfentanil. Patient variables such as age, sex, weight and height may need to be inputted to complete the algorithm. The algorithm, using this information together with the known drug concentration in the syringe, thereafter allows the anaesthetist to simply select a target patient concentration of drug. The pump automatically alters the infusion rate to most rapidly attain and maintain the calculated drug concentration in the patient at the set target. Such systems are very simple to use and obviate the need for complex calculations and detailed knowledge of pharmacokinetics by the user. They are largely responsible for a revolution in the administration of anaesthesia in a trend away from volatile agent maintenance towards total intravenous anaesthesia (TIVA).

Depending on the design of the chip these functionalities may be imbedded as hardware or firmware or added as software instructions. It can be seen how one basic infusion pump chassis can thus be made to give rise to effectively many different devices according to the microprocessor configuration. Each configuration or mode of use has its own attendant hazards in addition to those general to infusion pumps. Although all functionalities may be available in the one device, for risk management they may be selectively disabled through a restricted access menu such that only facilities that are needed in a clinical area are available to the user.

Modern devices are also able to communicate with centralized automated data archiving systems to automatically record the administration rate of a drug alongside the output of patient physiological monitors.

Simple infusion systems

In the operating theatre where there is closer observation of the patient’s hydration and circulating volume as necessitated by their rapidly changing status, most intravenous fluids are still administered under gravity from flexible plastic containers using single-use fluid administration sets. These are of several types:

• simple fluid administration, no filter, droplet size approximately 15 drops/ml (0.07 ml)

• blood and fluid administration with clot filter at about 200 μm mesh size. These giving sets use a larger-bore tubing and a double drip chamber containing a float: by squeezing the bottom chamber, the float jams the inlet and acts as a one-way ball valve allowing fluid to be pumped. Droplet size is usually approximately 15 drops/ml

• burette, 100–150 ml in volume, droplet size either 15 drops/ml or, for paediatric usage, 60 drops/ml. A flap or ball valve at the bottom of the burette prevents air entering the drip chamber when the burette is empty

• platelet giving sets, designed to reduce the risk of aggregation in the giving set as would occur with conventional blood giving sets.

Some giving sets are now also designed so as to also be compatible with volumetric infusion pumps. This necessitates a narrower-bore tube made from softer plastic to function with the peristaltic pumps (see below). Flow rates are, therefore, lower and the tubing is less kink-resistant. These sets usually have a 15 μm filter at the base of the drip chamber and are not suitable for infusing blood or for use in adult resuscitation.

The rate of infusion in a simple gravity fed system depends on:

• the height of the fluid container above the infusion site

• the resistance to flow caused by the infusion set

• occlusion of the tubing from a rate controlling device

• the physical properties (i.e. viscosity) of the fluid to be administered

• the bore of the intravenous cannula

• the hydrostatic pressure in the veins of the patient.

The manufacturers of giving sets quote the size of drops as number of drops per millilitre, usually between 10 and 60, but it must be remembered that the actual volume of the drops depends on the physical properties of the fluid being administered.

Rapid infusion

When rapid infusion of blood or other fluids is required, the rate of administration may be increased by the use of an inflatable pressure bag (Fig. 19.3), which may even be contained in a rigid box to increase the speed at which pressure may be applied.

Figure 19.3 Pressure infusion device.

When fluids are administered at a rapid rate, provision should be made for warming them to body temperature, otherwise significant cooling of the body may ensue (see Chapter 30).

Principles of infusion devices

Pumped infusion systems overcome the variation in infusion rates caused by changes in back-pressure, tubing resistance and the vertical height of the fluid container above the patient. Hence, they have to be powered by some form of motor, which must be coupled to a mechanism for driving the fluid

The stepper motor

The driving force in the majority of infusion pumps and electronic syringe drivers is provided by an electronic stepper motor, which is directly controlled from a digital microprocessor system. The speed of a conventional electric motor driven from either an AC supply or a DC supply may vary with mechanical load, the voltage or the frequency of the supply. It is, therefore, difficult, without electronic feedback, to control such a motor accurately. The stepper motor (Fig. 19.4) is designed so that a series of pulses applied to the stator windings of the motor cause the shaft to rotate by a fixed amount for each pulse, typically 1.8°, 2.5°, 3.75°or 7.5°, irrespective (within certain limits) of the load. Infusion systems are designed so that a pulse generator, whose output frequency is varied by the microprocessor, can produce accurate control of an infusion by varying the speed of the stepper motor.

Figure 19.4 The stepper motor and lead screw assembly of an Alaris PK syringe pump during manufacture.

Originally, the most accurate (and expensive) infusion pumps used syringe type cassettes (see Fig. 19.2) and were referred to as volumetric infusion pumps. This, like the newer peristaltic pumps, is driven by a stepper motor controlled directly by a microprocessor. The volume of the cassette is typically about 5 ml, with the dedicated disposable ‘syringe cassette’ for each manufacturer’s pump being supplied separately as a sterile product. A valve operating in harmony with the piston directs flow from the infusate bag to the reservoir or from there to the patient. Fluid is drawn rapidly from the reservoir bag into the cassette in less than 1 s. The valve is then actuated such that on the piston upstroke the contents are expelled at the required rate into the patient, and the cycle is repeated. Although effectively this produces an intermittent flow, it also gives overall extremely accurate infusion rates, with only infrequent 1 s interruptions.

These are seen much less frequently now owing to the expense of the disposables and the feasibility of getting good accuracy using simple intravenous giving sets in modern peristaltic pumps.

Peristaltic pumps

The principle of the peristaltic infusion pump is shown in Fig. 19.5. The tubing of a giving set is compressed by a series of rotating rollers or by a wave of mechanical ‘fingers’ or cam followers. The section of tubing in the peristaltic mechanism must be hard wearing, of known and consistent internal volume, and have no memory after compression so that it easily fills on being released. Depending on the manufacturer, specific proprietary tubing with dedicated fittings may be needed to allow it to be loaded into the pumping mechanism. Precision silicone tubing is often used in this section of the giving set.

Figure 19.5 A. Linear, and B. rotary peristaltic mechanisms.

Rotary peristaltic pumps are now more often seen in use for the less demanding requirements of enteral feeding rather than intravenous administration.

Linear peristaltic mechanisms allow much easier loading of the infusion tubing and are now by far the commonest design of infusion pump. The driving force is again a stepper motor. The rotary motion from the motor is translated into a linear peristalsis by the use of cams and cam followers as shown in Fig. 19.6.

Figure 19.6 The rotary motion from the motor is translated into a linear peristalsis by the use of cams and cam followers.

Because such infusion pumps have the theoretical capacity to inject limitless quantities of air into a patient should air ingress occur upstream of the pump (for example due to an empty infusion bottle), these devices incorporate sophisticated ultrasonics (Fig. 19.7) or optics-based ‘air in line’ detection systems capable of sensing air bubbles as small as 0.1 ml volume. These are usually placed downstream of the pump mechanism. Further protection is conferred by setting target delivery volumes smaller than the volume in the bag of infusate.

Figure 19.7 Braun Infusomat Space infusion pump with front cover opened to show (1) line pressure sensor, (2) ultrasonic air sensor, (3) peristaltic mechanism with precision silicone tubing loaded.

To detect obstructed or extravasated catheters, electrically powered infusion devices must have some measure of the pressure generated in the infusion line beyond the device. Peristaltic intravenous pumps, therefore, use a sensing piston pressing on the infusion line immediately downstream of the pumping chamber. This is calibrated to indirectly measure line pressure and can be programmed to alarm for occlusion at different pre-set levels.

Syringe drivers

There continues to be a range of small simple battery-operated syringe drivers (Fig. 19.8). The driving mechanism is a miniature DC motor that is switched on and off intermittently and drives a screw-threaded rod (lead screw), which is linked to the syringe plunger, causing its advancement. They may have a variable rate that is altered by adjusting a recessed control using a small screwdriver. These pumps are small and light enough to be worn in a holster by an ambulant patient and are now used chiefly for narcotic infusions for the relief of cancer pain. Great care must be taken in calculating drug dilutions and to ascertain that the correct units are used for setting the infusion rates, as the pumps are available in different models with rates set either as mm per 24 h or mm per h of plunger movement (Fig. 19.9).

Figure 19.8 Graseby syringe driver.

Figure 19.9 Graseby battery operated syringe drivers calibrated in mm per h and mm per 24 h.

Photo courtesy M Stewart, Graseby Medical Ltd.

Virtually all other syringe drivers for hospital use and particularly those used in intensive care and anaesthesia use microprocessor controlled stepper motors, again connected to the syringe plunger by a carriage on a lead screw (Fig. 19.10). Thus, each pulse applied to the stepper motor causes the advancement of the syringe plunger by a known amount. The pulse generator may be calibrated from 0.1 ml h⁻¹ to 1200 or occasionally 1800 ml h⁻¹, the higher rates being used only for delivering a bolus (often of predetermined volume) or for purging the infusion line.

Figure 19.10 Syringe driver with microprocessor-controlled stepper motor, connected to the syringe plunger by a carriage on a lead screw.

Syringe pumps (the term is synonymous with syringe drivers) are now designed to automatically recognize a variety of syringes by virtue of the calibre of the barrel using some form of spring-loaded arm; some manufacturers’ models nonetheless require manual confirmation of the detector. Infusion line pressure (and empty syringe detection) is calculated indirectly from the force acting on the syringe plunger by sensors, which may be incorporated into the carriage or lead screw assembly (Fig. 19.11). This is a more popular option than the use of specialized infusion sets with in-built diaphragm and corresponding transducer housing on the syringe pump which remains largely confined to pumps used on neonatal ICUs. The facility in some devices to also alarm for low infusion line pressures is intended to allow recognition of disconnection of an infusion line (with the aim of, for example, preventing awareness in intravenous anaesthesia).

Figure 19.11 A piezoresistor is bonded onto the metal chassis which carries the lead screw of an Alaris Asena syringe pump. The distortion (invisible to the eye) of the steel plate caused by the force acting on the lead screw causes a change in resistance which can be calibrated to be read as a pressure within the syringe and giving set.

Rechargeable batteries

Although mains-driven, electrical infusion devices must have battery back-up both to cover mains failure and for patient transfer and emergency situations. The performance of the in-built rechargeable batteries is an important consideration when purchasing such equipment, but it must be remembered that this is also influenced by the battery maintenance procedures. Poor battery life can render otherwise excellent devices unreliable and unusable. Pumps should be kept connected to the mains when not in use and batteries should be replaced appropriately. Microprocessor-driven infusion devices are susceptible to bizarre error conditions when rechargeable batteries begin to fail.

In common with many rechargeable batteries, nickel-cadmium (NiCd) rechargeable batteries should be periodically run down completely to prevent the development of ‘memory’, which renders them unable to discharge their full capacity. NiCd batteries are gradually being replaced by nickel-metal hydride (NiMH) batteries for environmental reasons (cadmium is a toxic heavy metal). In comparison to NiCd batteries, NiMH batteries have a higher-energy density, i.e. they can hold more charge per unit weight, but have a more limited service life. They are similarly prone to memory problems and need appropriate maintenance. Lead acid batteries, also called ‘sealed lead acid’ to designate portability and to differentiate from the flooded type used in cars, have no ‘memory’, are cheap and reliable, but have long charging times. They are most often found on portable equipment such as ventilators and other heavy devices (e.g. wheelchairs and ‘uninterruptible power supply’ systems). Lead acid batteries, conversely, suffer by being allowed to fully discharge and must not be stored in this state as the process of sulphation can render them unusable. Lithium-ion (and lithium polymer) batteries have a high-energy density and no memory but are very expensive. The technology is currently confined largely to portable personal electronic equipment such as mobile telephones.

Battery maintenance is difficult in the hospital setting. Medical device batteries obviously cannot be safely run down whilst in use. Similarly, encouraging even partial discharge of a battery decreases the safety margins in the event of power failure or the need for transportation of a device.

Safety

Microprocessor-driven infusion devices are used for the administration of many potent drugs with narrow therapeutic windows where maladministration can have lethal consequences. The drugs are used in a variety of dilutions and dosed in units that can vary by several orders of magnitude (ηg/ml, µg/ml, µg/kg/min, mg/kg/h). The devices themselves are highly versatile and capable of being instructed to perform in many different manners. Given these factors it is evident that there is significant potential for user error with disastrous consequences.

The devices themselves may also malfunction, albeit infrequently. The pump processor monitors many aspects of the device’s performance in order to detect malfunction and to make operation safe. Though rare, glitches in the software may under certain circumstances cause over- or under-infusion. Software issues are more commonly seen as a stopped infusion when the processor receives apparently conflicting messages from different sources in the device, which then is made to fail safe with appropriate alarms and error codes. An unwanted termination of infusion can be equally dangerous – if, for example, the patient’s circulation is dependent on vasoactive drugs.

The machines have built-in alarms for occlusion, low battery, mains failure, disengagement of drive mechanism, failure to load infusion set and other common fault conditions. In spite of this, they remain high-risk devices capable ultimately of delivering drugs dangerously: they have a recognized associated morbidity and mortality. In at least 27% of the 1495 incidents involving infusion pumps reported to the Medical Devices Agency in the UK between 1990 and 2000, the cause was found to be user error (including failure to maintain the device appropriately).¹ Only in 20%, were problems device-related with issues such as performance, degradation, quality assurance and design and labelling. In the 53% of cases where no cause was established it is likely that a very large number represent user error. The MHRA report for the year 2009 again reports that infusion and feeding pumps ‘continues to be one of our busiest areas’ with 375 adverse incident reports.² The user must, therefore, be ever vigilant and particularly aware of the following problems:

• Unitary programming errors. The simplest and most common error is a mistake or slip in selecting the correct dosing unit or drug concentration, e.g. mg/kg/min instead of µg/kg/min or µg/ml instead of mg/ml. These issues are addressed to a large extent by the use of drug libraries (v.i.).

• Wrong drug errors. Where more than one infusion pump is used particularly when they are controlled through a common interface it is relatively easy to confuse the drugs. Propofol available as identical looking 2% and 1% solutions is also easy to confuse especially as one manufacturer of open label TCI pumps allows a change in concentration during a TCI (Fresenius) and one does not (Alaris), and hence does not prompt for concentration when new syringes are loaded. The Diprifusor system automatically senses the drug in the prefilled syringe and hence does not have this problem but this may now predispose doctors to errors when using other TCI systems.

• Failure to restart infusion. This is a very common error after refilling syringes during intravenous anaesthesia. Software changes across generations of the same device may require additional key presses for the same function (Diprifusor) and can contribute further to this error.

• Siphoning. This is the term used to describe the uncontrolled flow of fluid from a syringe into the patient under gravity. Modern drivers clamp the plunger and have a detector or mechanism built in to protect against incorrectly mounting the syringe plunger. It is best, even with modern designs, to not have the syringe driver higher than the patient, as small amounts of siphoning can still occur. Anti-siphon valves – essentially a one-way valve with a high opening pressure – may be incorporated into syringe pump giving sets (see below, Infusion lines). The possibility of inadvertent administration with infusion pumps as a result of administration sets not being properly clamped off when the pump is not in operation or when the infusion set has not loaded into the device properly still exists but should be largely designed out in new devices.

• Reflux. Where multiple infusion lines are connected to a single intravascular device and there is a distal obstruction it is possible for drugs from pumped lines to reflux up an attached gravity fed line. It is imperative that such lines have anti-reflux valves incorporated (see below, Infusion lines). It is worth stating that this fault situation can and does also frequently occur when drugs are injected into side ports of intravascular catheters or simple giving sets administering fluids by gravity feed. (Classically muscle relaxants injected into giving sets with blocked intravascular catheters, which are then cleared by postoperative recovery staff, resulting in a paralyzed awake patient.)

• Common deadspace and multiple infusions. See ‘Infusion lines’ below.

• Infusion of air. More commonly now due to unprimed infusion lines.

• Extravasation injuries. Continued infusion after a cannula has ‘tissued’ or become extravascular are not prevented by the use of low infusion line pressure limits.

• Electrical hazards. These are present as with all powered devices, but the common liquid spills on to infusion pumps can be particularly dangerous in the presence of cracked machine casings or frayed mains electricity leads.

• Software revisions in microprocessor-driven devices. These can produce a wholly new machine within the familiar appearance of the old. Although features may be added or improved, there is also the possibility of introducing new problems and errors, particularly for those familiar with previous versions. Manufacturers should treat all but the most trivial software revisions as new devices and issue new instruction and training/maintenance manuals.

Because of the huge flexibility of microprocessor-controlled infusion devices, it is increasingly important for users of these devices to have familiarized themselves specifically with the features and functions of each model before clinical application. Programming errors are very common and may potentially result in lethal overdosage. These devices are not always entirely intuitive to use, errors are commonplace with, for example, the ‘hands free bolus’ facility (a pre-programmable bolus dose that does not require the button to be kept depressed during delivery), which may give the option of a variety of unexpected units and infusion rates.

Image courtesy of CareFusion.

Line pressure limits

The line pressure limit at which devices alarm for occlusion can usually be altered; for some models this is a user function whilst on others it must be performed by biomedical engineering departments. Limits must be set such that nuisance alarms do not occur due to the resistance of the giving set or ‘stiction’ at the syringe barrel/plunger interface. High-pressure limits, however, have two major disadvantages:

1. It will take longer to alarm for occlusion (alarm time is increased) particularly at low infusion rates and at the start of an infusion – most commonly due to leaving a stopcock closed.

2. On release of the obstruction a proportionately larger bolus will be delivered to the patient; this may be clinically significant.

Electromagnetic interference

Microprocessors are susceptible to electromagnetic interference (EMI) from wireless telecommunication devices such as mobile telephones. EMI can cause serious malfunction in medical devices. The strength of the electromagnetic field is proportional to the power output of the telecommunication device and inversely proportional to the square of the distance from the source. The possibility of EMI is greatest with emergency radios (as used by ambulance, police and fire services) followed by security radios (as used by hospital maintenance and portering staff) and, finally, by mobile phones (both analogue and digital). The Medical Devices Agency of the UK reports 41% of the medical devices they tested suffered interference from emergency radio handsets at a distance of 1 m with 49% of these being classed as serious. By comparison, the figures for security radio handsets were 35% and 49%, respectively, and those for mobile phones 4% and 0.1%.³ It is important to note that these devices can cause interference even in standby mode and must, therefore, be fully switched off to be considered safe. Cordless telephones and wireless computer local area networks do not appear to cause significant interference.

Although most hospitals have policies demanding that mobile phones be switched off in clinical areas, clinicians must always bear in mind the potential for such malfunction, given the ubiquitous nature of mobile telephones and the increased risk of problems with the two-way radios used in hospitals.

Target-controlled infusion (tci)

Intravenous anaesthesia, referring to anaesthesia maintained by continuous infusion of anaesthetic agents, requires that consistent drug concentrations can be achieved which may be altered in response to the needs of the procedure. The term TIVA is used synonymously with intravenously maintained anaesthesia, but perhaps should be reserved for those scenarios where no inhalational agents (including nitrous oxide) are co-administered.

Although the drugs used for intravenously maintained anaesthesia are of rapid onset and have short half-lives and durations of action, in order to rapidly achieve and maintain a given clinical effect it is still necessary with currently available agents to administer boluses and then to reduce progressively the infusion rate. TCI describes a system whereby a computer controls the rate of infusion of a drug to achieve (in as short a time as possible) and maintain any given target concentration. In use this means that the anaesthetist is relieved of having to continuously make complex calculations and adjustments of the infusion rate, thus lending the ability to endlessly vary the drug concentration to achieve the desired effect (Fig. 19.14).

Figure 19.14 Relationships in a target-controlled infusion system.

Investigators in the fields of anaesthesia and pharmaco-kinetics have considerable experience in programming computers to achieve constant concentrations of drug by driving infusion pumps at variable rates. Many different terms have been used to describe this process since the description of the bolus elimination and transfer (BET) infusion scheme with a system called CATIA (computer assisted total intravenous anaesthesia) in 1983.⁴ The aim throughout has been to emulate the simplicity of administration of inhalational agents, in effect to design a ‘calibrated vaporizer’ for intravenous agents, i.e. a system akin to that for inhalational agents wherein a given setting on the ‘vaporizer’ will, within certain limits, result in a similar and constant drug concentration in the patient.

Although the original experimental systems were cumbersome, in 1990 workers from the University of Glasgow described a system for the delivery of propofol using a conventional syringe pump driven by a Psion II hand-held computer.⁵ There was good correlation between measured and predicted plasma concentrations, and users found the device easy to use. This was the basis for the Diprifusor system, for many years the only commercial system for delivering a drug using a pharmacokinetic algorithm. Since 2005 various device manufacturers have incorporated TCI algorithms into their infusion pumps, these are examined under ‘open label TCI’.

Key components

A TCI system must have the following key components:

• an infusion device

• a computer, and a pharmacokinetic algorithm for each drug which translates predictions from the model into instructions to control the infusion device, the algorithm must have the correct pharmacokinetic data set (constants) for the patient group in which it will be used

• patient-specific data: usually weight (the Diprifusor system does not alter the data set for patient age or circumstance); some algorithms may need other co-parameters such as age, sex and height

• a user interface to allow and confirm data input and to inform the user of the algorithm’s predictions

• a fail-safe system: a safety mechanism to shut down the system in the event of computer failure; this may be in the form of a second microprocessor calculating the drug concentration based upon the output of the pump and comparing this with the predicted value from the first microprocessor’s control algorithm, with instructions to shut down the system if there is greater than a preset discrepancy (5% for Diprifusor).

It is important that the algorithm is not inadvertently reset as if for a new patient when renewing drug syringes during the course of an anaesthetic, as this will of course result in the delivery of a further loading dose.

Diprifusor

In 1996 the manufacturer Graseby produced anaesthesia syringe drivers bearing the ‘Diprifusor’ TCI Subsystem (Fig. 19.15) licensed from Zeneca (now AstraZeneca), the makers of propofol. The Diprifusor chip has subsequently been incorporated into other makes of syringe driver, the devices being readily identified by the associated green and white logo. An advantage of this modular approach to the design of TCI systems, where the pharmaceutical company retains responsibility for the Diprifusor module, is the standardization of propofol delivery, such that all pumps bearing that module will deliver the same amount of drug for any given target setting because the same pharmacokinetic model is operating (cf. open label TCI issues). To date, in excess of 20 000 such modules (Fig. 19.16) have been sold worldwide (personal communication, AstraZeneca). The Subsystem comprises components designed to:

Figure 19.15 Graseby 3500 Anaesthesia Pump incorporating the Diprifusor chip and demonstrating the associated logo.

Figure 19.16 A. The Diprifusor subsystem installed on the microprocessor control board of the Graseby 3500. B. Note the ‘aerial’ (A) for recognition of the syringe tag.

• recognize electronically tagged pre-filled syringes of propofol produced by AstraZeneca in order to identify drug and concentration and enable TCI mode when appropriate

• control syringe pump in TCI mode to deliver propofol by TCI.

In order to ensure that the syringe driver operates in TCI mode only when loaded with the correct drug (proprietary propofol as Diprivan), the pre-filled syringes carry a small electronic tag in the finger grip of the barrel. When the syringe is correctly located the ‘Programmable Magnetic Resonance’ tag lies within a recess close to an aerial in the body of the pump. In response to signals from the aerial the magnets within the tag oscillate at particular frequencies generating a signal, which is picked up by the aerial. To discourage refilling of the syringe the tag is subsequently ‘wiped’ by the pump when the actuator pushing the syringe plunger reaches a pre-set travel.

The pharmacokinetic model used is based on the Marsh open three-compartment model ⁶ (Fig. 19.17) with parameters as shown in Table 19.1. Fig. 19.18 shows a typical infusion scheme from a TCI device to demonstrate the bolus and decreasing infusion rate needed to maintain an increased concentration. Diprifusor software (which targets the plasma concentration as the control variable) also allows the predicted effect site concentration to be demonstrated at any time. This facility was added slightly later using a rate constant for equilibration of the effect site (k_e0) of 0.26 min⁻¹ derived from a separate study.⁷ Effect site display is useful, both in terms of demonstrating for impatient anaesthetists the magnitude of the time lag between the compartments and for correlating clinical effect with theoretical concentration when plasma compartment ‘overpressure’ is used to speed up the onset of effect.

Figure 19.17 Schematic representation of open three-compartment model.

Table 19.1 Pharmacokinetic parameters for propofol incorporated in Diprifusor software

V₁ Volume of central compartment	228 ml kg⁻¹
Half life of delay central to effect site (t_1/2Keo)	2.6 min
Rate constants:
K₁₀ (elimination rate constant)	0.119 min⁻¹
K₁₂	0.114 min⁻¹
K₂₁	0.055 min⁻¹
K₁₃	0.042 min⁻¹
K₃₁	0.003 min⁻¹

Figure 19.18 A typical infusion scheme from a target-controlled infusion device, showing the predicted plasma concentration in response to alterations in the set target and the infusion rate.

Accuracy

Two independent factors contribute to the overall accuracy of a TCI system:

1. Delivery performance of the infusion system: how well the syringe pump outputs the desired volume of drug

2. Predictive accuracy of the pharmacokinetic model.

Syringe pumps incorporating Diprifusor are required by the manufacturer to have a performance such that the infusion error at specific time points is within ±5% of the ideal volume.⁸

In considering figures for the predictive accuracy of pharmacokinetic models, a number of issues must be borne in mind:

• inter-patient pharmacokinetic variability must always be expected and the need to titrate drug target concentration to achieve the desired effect is inescapable

• performance is likely to vary with time and chosen target concentrations

• errors in blood sampling technique or variability in the measurement of blood concentrations; this may at times be due to inadequate mixing at high infusion rates

• in clinical assessments of accuracy the co-administration of other drugs may affect the handling of the target drug.

There is an attempt to standardize measurement of the accuracy of computer-controlled infusion systems.⁹ Performance error expressed as a percentage is calculated as shown below:

where C_M and C_CALC are the measured and calculated blood drug concentrations. Other indices are derived from this. The median performance error (MDPE) is a measure of the tendency of the system to over- or underestimate the measured blood concentration in a given patient or scenario. This is the degree of bias and has direction as well as value, thus if bias has a positive value the measured concentrations tend to be greater than predicted. The precision, or size, of the error is represented by the median absolute performance error (MDAPE). Divergence and wobble reflect on time-related changes in performance and intra-subject variability in performance respectively. Typical figures for MDPE of 16.2% and MDAPE of 24.1% were reported for Diprifusor in one study,¹⁰ with measured concentrations tending to be higher, particularly after induction or an increase in target concentration and at higher targets. In context these figures compare favourably with the performance of inhalational anaesthesia, where after 15 min of isoflurane administration a ratio of 0.78 is reported between arterial and end-tidal partial pressures¹¹ with the concentration remaining 20% lower at 1 h. In this context there is an even greater difference between vaporizer setting and arterial concentration.¹²

Open label TCI

The expiry of the patent on propofol and the general acceptance of TCI as a mode of drug delivery have driven other manufacturers to incorporate TCI algorithms into their syringe drivers. These are known as ‘open label protocols’ because they do not require the use of the tagged syringes of Diprivan. At the same time the popularity of remifentanil as the opioid agent in TIVA naturally led to demands for an appropriate TCI algorithm for that drug also. The resulting devices have been approved under medical device legislation without any reference to the pharmaceutical company responsible for drug labelling, although the joint role of some manufacturers as being also producers of the drug propofol does appear to constrain their readiness to offer further models for propofol.

Currently in the UK, open label pumps are available from four manufacturers. All offer a choice of Schnider ¹³ and Marsh pharmacokinetic (PK) models to drive the propofol TCI and the same single model (Minto¹⁴) for TCI remifentanil.

TCI sufentanil using the Gepts ¹⁵ model is also available from all the device manufacturers mentioned here except B Braun. At the time of writing, two companies offer further drug models. CareFusion in the Alaris PK pump also offer paediatric propofol pharmacokinetic models – Kataria¹⁶ and Paedfusor¹⁷ – and a model for alfentanil (Maitre¹⁸). Arcomed AG (Switzerland) can offer two versions of the Marsh model as well as Kataria and Maitre models and have expressed a willingness to consider implementing other models (with due licensing disclaimers) at customers’ specific requests (personal communication). B Braun and Arcomed AG are able to implement the TCI algorithms into their peristaltic infusion pumps as well as the more traditional syringe pumps, adding another degree of flexibility.

Where the pharmacokinetic (PK) models have the facility to calculate the effect site concentration the new open label pumps can offer the option of using the calculated effect site, rather than the plasma concentration, as the control variable for driving the infusion. Remifentanil effect site control using the Minto model is offered by all devices, similarly propofol effect site control on the Schnider model. Elsewhere calculations and control of effect site concentration have introduced a whole new area of confusion for users of open label pumps. The reasons for this, namely the methodological details of pharmacokinetic/pharmacodynamic (PK/PD) modelling and attendant controversies, are well beyond the remit of a text on equipment, but the consequences are relevant, as described below.

Ultimately the various devices all offer much the same functionality whether they are controlled through one interface as in the Orchestra Base Primea from Fresenius (Fig. 19.19) or programmed and controlled individually as in the Alaris (Fig. 19.20) and B Braun (Fig. 19.21) devices. For most prospective users the choice between them is probably more a matter of balancing personal preference for the input method and display against the irritating interface and programming quirks. The devices all suffer from not having a numeric keyboard – thus having to scroll through numbers for entering values – and needing preloading with infusion schemes to allow use in anything other than TCI mode or simple ml/h control.

Figure 19.19 Orchestra Base Primea (Fresenius Kabi AG) allows simultaneous control of several devices through one control panel (screen display shown here as inlay).

Figure 19.20 Alaris PK pump.

Photograph courtesy of Carefusion.

Figure 19.21 Perfusor Space, B Braun.

The potential for confusion and error on the whole has been markedly increased. This is as a consequence of:

• having two or more pumps side by side with no automatic drug identification

• differing user interfaces and set up procedures between devices

• issues with pharmacokinetic models: their limitations and varied implementations by manufacturers, as below.

Other pharmacokinetic models

It is paramount for anaesthetists to recognize that a given target concentration obtained using one particular model (for example the Marsh model as used by Diprifusor) does not necessarily translate to the same clinical effect when that target is delivered by a different PK model (for example, the Schnider model as available in the open label TCI pumps). This is because it is precisely that which causes the models to be different (i.e. the predicted concentration following a given dose) that will cause them to give differing amounts of drug for a given target concentration. Clinical effect is clearly a function of dose of drug given and, hence, differing effects must be expected when using different models targeting the same concentration.

Additional confusion surrounds the issue of effect site control of the infusion, with multiple models and manufacturers updating the Marsh model with their own interpretations of available data (Table 19.2). The rate constant (k_e0) that was originally added to Diprifusor to calculate and display the effect site concentration of propofol was felt by some to be too slow, but because of the predominance of Diprifusor it is a figure that has clinical relevance for many and is used elsewhere. Fresenius uses a Marsh model with a much a larger (k_e0) for effect site control, but this is felt now to be too fast.¹⁹ Arcomed offer two versions of Marsh for effect site control, each effectively using figures at opposite ends of the spectrum. The Alaris PK uses the same value as Diprifusor, and Braun use an intermediate figure for their Marsh model, but neither currently allows effect site TCI using the Marsh model.

Table 19.2 The differing rate constants for equilibration of the effect site with the plasma propofol concentration (also expressed as half time) used by different manufacturers

Thus, in effect site targeting mode a given target concentration can result in different volumes of drug being delivered from different manufacturers even when nominally using the same model (see Fig. 19.30). Even for those with an interest in these issues, it is difficult to learn or remember how a PK model and device combination will behave under different circumstances, particularly once patient variability is also considered!

Extremes of weight also present a problem for the current PK models. This is most notable in the obese, where the James equations ²⁰ used by the Schnider and Minto models in the calculation of lean body mass (LBM) for their algorithms, generate decreasing values above a certain body weight depending on height and gender. As a result Fresenius pumps will not progress with TCI for body mass index (BMI) >42 for males and >35 for females, these being the inflection points on the James equation graphs after which the LBM starts to decrease. Alaris pumps handle the problem differently; they simply will not allow a greater weight entry then the corresponding figure for the above BMIs. The Marsh model (and hence Diprifusor) markedly underestimates plasma concentrations for morbidly obese patients (i.e. delivers too much drug for a given concentration) and the patient weight used in the algorithm needs to be manipulated significantly towards the ideal body weight of 70 kg for both the obese and the underweight.

Future developments

Open TCI initiative

It can be seen from the problems above that great confusion can arise from the existence of the different models and their various implementations by the manufacturers. Significant detailed knowledge of the models and manufacturers’ implementations of them is needed if anaesthetists are not to be surprised by these devices in use.

The ‘Open TCI initiative’ is an attempt by workers in the field of drug pharmacokinetics and pharmacodynamics to come up with a single unifying model for each drug on which all workers are agreed and which would be the basis for all dosing by TCI.¹⁹ Manufacturers of medical devices would, thus, not be in the position of having to dictate drug dosing. It will also become possible to compare effects and outcomes between studies once drug delivery is based on a common dosing regimen.

Experimental systems

A logical development from the TCI system, for maintaining constant concentrations of anaesthetic, has been to link this with some form of depth of anaesthesia monitor to automatically alter the chosen target to maintain a constant depth of anaesthesia. Closed loop anaesthesia (CLAN) has been successfully used to anaesthetize patients breathing spontaneously during surgery.^21,²²

TCI may also be used as the basis for an advanced form of PCA (or sedation). Here, the patient’s button presses may, for example, result in an increase or prevent a decrease in the target concentration of drug rather than a drug bolus.²³ Such systems are so far experimental only.

Patient-controlled analgesia (pca)

The quality of postoperative pain relief and patients’ sense of autonomy have perhaps improved with the use of PCA. Although intramuscular bolusing with opiates at 3–4 hourly intervals may provide good analgesia, the reluctance of patients to disturb hard-pressed nurses and the need for the increasingly sparse presence of a second member of staff to administer ‘controlled drugs’ rarely results in optimal analgesia. The advantages of PCA are:

• patient autonomy

• rapid relief of pain

• analgesia/dosage tailored to patient’s requirements, with the patient able to balance analgesia and side-effects.

For safety reasons most physicians are reluctant to use a background infusion alongside the patient-controlled bolus doses. This is the key disadvantage: bolus doses are necessarily small with a short duration of action resulting again in fluctuating analgesia levels, particularly at night time where patients often complain of waking in pain and thus poor sleep.

Key points in any PCA system are:

PCA is most commonly delivered intravenously, but may also be subcutaneous or into the epidural space, although epidural PCA does have the theoretical problem of bypassing the safety feedback loop seen with intravenous opiates, where the potentially overdosed patient becomes somnolent and unable to make further analgesic demands. The simplest method of administration is by bolus dosing; here a preset bolus, which may be programmed either by volume or in more complex systems by weight of drug in milligrams, is injected on demand. There is a pre-programmed lock-out time during which further demands are ignored, thus allowing for the time of onset of action of the previous bolus and protecting against overdose.

A background continuous infusion may also be available with some devices. This militates against patients waking in pain due to protracted periods of no analgesic administration, although safety considerations make this a rarely used facility (see above).

PCA devices (Fig. 19.22) are mostly variations of the infusion devices discussed earlier. They may be syringe drivers or peristaltic type pumps, which are microprocessor-controlled; hence, again the aspects of those devices mentioned elsewhere are still pertinent. Because of the trend towards increased patient mobility and patient-controlled epidural analgesia (PCEA), a number of manufacturers now produce miniaturized PCA pumps, often using a small compressible pumping chamber designed as an integral part of the giving set (Fig. 19.23). Size and portability are limited by the size of the drug reservoir and the need to encase this in a lockable anti-tamper shell.

Figure 19.22 A Graseby PCA.

Figure 19.23 A Baxter PCA pump which may be used for epidural continuous infusion or even epidural PCA.

As with other microprocessor-controlled devices, there is a vast amount of information recorded by these newer devices, only some of which is presented to the clinician. This includes times of error messages, alarm conditions, ‘power up’ and ‘power down’ and rate or programme changes. There is usually a facility to view and/or download or print the ‘patient history’ so that pump settings may be tailored to the patient’s needs. A further feature is the lockable control panel (this may be an actual key and lock in some machines) needing access codes and button presses of unlabelled ‘soft-keys’ to prevent unauthorized tampering with the pumps.

A consequence of increasing functionality, together with miniaturization and the need for security, is that many of these devices are no longer intuitive to use and effectively need a dedicated ‘pilot’. Disenfranchised staff become reluctant or unable to initiate and alter settings with the quality of analgesia suffering as a result. Uniformity of equipment in clinical areas helps, but when deciding a purchasing policy, the considerable capital costs of these devices, the cost of initiation and maintenance of staff training, and the sheer numbers of devices needed to provide the service are also to be considered.

Elastomeric pumps

Given the issues surrounding microprocessor-controlled infusion pumps and the specific problems associated with the PCA configurations, there is much to be said for the use of simple low-cost disposable PCA devices, as shown in Fig. 19.24.

Figure 19.24 An elastomeric disposable PCA pump with the demand button worn as a ‘wrist watch’.

Elastomeric devices are in effect powered by the energy stored in the stretching of the balloon holding the reservoir of drug. A small compressible chamber of preset volume (usually 0.5 or 1 ml), on a wrist strap or attached to the reservoir, holds the bolus dose which is delivered to the patient through a one-way valve when the chamber is squeezed. The lock-out time is predetermined by a flow restrictor connecting the reservoir and chamber, which governs the rate of refilling of the chamber, thus giving a maximal flow rate through the device, typically 5 ml h⁻¹. Drug demands, before the effective lock-out time is reached, result in proportionately smaller doses. PCA regimen alterations are, therefore, made by altering the drug concentration in the reservoir. Some newer models have a user (prescriber) variable flow restrictor and or bolus size (Fig. 19.25).

Figure 19.25 ON-Q elastomeric pump with variable rate infusion controller and bolus device.

Photograph courtesy of I-Flow Corporation, Lake Forest, USA.

Additionally, all such elastomeric devices if appropriately configured, can just as easily be used for the infusion of local anaesthetics either into the operative site or around major nerves for regional anaesthesia. For those devices designed for local anaesthetic infusion, the flow restrictor may be the catheter itself by virtue of its bore and length, with some systems allowing multiple such catheters to be run from one reservoir (e.g. Alpha range of pumps and catheters from Advanced Infusion, San Dimas, USA).

Flow accuracy for elastomeric devices is typically within 10–20% of the given rate, but depending on the design of the device, over- or under-filling of the reservoir can affect flow rate. Flow rate is also affected by temperature, particularly at the flow restrictor, and it is important that manufacturers’ instructions are followed as to whether, for example, the tubing and restrictor are placed close to the patient’s skin or worn outside the clothing. A change of 10°C in the temperature of water-based fluids results in altered viscosity causing a 20–30% change in flow rate.

In spite of these limitations, these devices are preferred by patients and nurses for their ease of use and, possibly, for the lack of alarms.

Of prime importance with PCA is safety and security. All devices must be safe against over-dosage, either caused by fault conditions or drug siphoning. There must also be security against tampering and adjustment of settings by unauthorized staff, patients or visitors.

Other non-electrically powered infusion devices

Gas- and spring-powered devices do exist, particularly for specialist applications such as implanted drug-delivery systems. They are not discussed further here and are not in common usage for PCA or routine postoperative ambulatory analgesia.

Autotransfusion

When a large blood loss may be expected during surgery, it may be possible to reuse the patient’s own blood, provided that it is uncontaminated and free from clots. Uncontaminated blood is aspirated by the machine, red cells are washed, spun down and suspended in crystalloid solution with added anticoagulant and stored temporarily in a reservoir. It may then be transfused back into the patient via a filter. The operation of such machines needs a dedicated member of staff.

Related equipment

Filtration

Intravenous fluids should be filtered to protect the patient from microscopic foreign material. Blood for transfusion that is more than 24 h old should be filtered (Fig. 19.26) to remove micro-aggregates that form from the breakdown products of the cellular components and platelets. Blood filters are of two basic forms: screen filters and depth filters. Screen filters function as ‘sieves’ and are usually constructed from a woven mesh. They have a regular pore size, often of 40 µm. The efficiency of a screen-type filter at removing foreign matter increases progressively with each unit of blood passed as the pore size tends to decrease progressively down to 20 µm. Screen-type filters are said to be less damaging to the red cells than are depth filters. Depth filters consist of a pack of synthetic fibre, often Dacron, not formed into a mesh. The mechanism of filtration is actually by adsorption of unwanted material, down to a size of about 10 µm onto the surface of the fibres. This adsorption is probably due to electrical charge differences between the particles and the fibres. With the depth filter, the efficiency at removal of unwanted material decreases with each unit of blood, probably as a result of channelling in the pack of fibres.

Figure 19.26 A blood filter.

Intravenous crystalloids may be filtered with much finer sieve-type filters to remove foreign particulate matter, including bacteria. The Pall intravenous ‘Site Saver’ extends the life of the giving set, which should normally be replaced every 24 h, to up to 96 h. The construction of the ‘Site Saver’ is shown in Fig. 19.27. All particulate matter larger than 0.2 µm is removed by the 0.2 µm filter membrane. Air can be vented through a hydrophobic membrane, which has 0.02 µm pores.

Figure 19.27 Pall intravenous ‘Site Saver’.

Infusion lines

When more than one fluid or drug is infused through a common intravascular device, alterations in the rate of infusion of one of these substances will temporarily affect the rate of administration of the remainder. The degree of perturbation is a function of the shared volume (dead-space) of intravenous device and the infusion rates (Fig. 19.28). For drugs with short half-lives this can have a dramatic effect. Ideally, each intravenous infusion of drug would be given through a dedicated infusion line, so that alterations in the rate of infusion of one drug do not affect the others. For simplicity and patient comfort, it is preferable to insert only one intravenous catheter and use a many-tailed infusion set. Such infusion sets are available specifically for TIVA with one or two long narrow calibre ‘pump lines’ and one wider bore line to connect to a giving set for fluid administration under gravity (Fig. 19.29). Each line should incorporate a one-way or anti-reflux valve to prevent backflow as well as anti-siphon valves on the pump lines. The shared volume, where the lines join before the intravenous catheter, should be as small as possible. Because such lines are made up of a number of components assembled together it should be borne in mind that they may leak or be obstructed at the bonded junctions. This may not always be obvious.