1 Introduction

Protein binding of drugs in humans is a well described phenomenon. Current regulatory processes in most countries make it compulsory for pharmaceutical companies to perform various studies for new drugs, including a detailed in vivo pharmacokinetic characterization. Data on parameters such as drug clearance, apparent volume of distribution (Vd) and protein binding are then used to develop dosing regimens which aim to ensure consistent achievement of target drug concentrations in the patient. The goal of this process is to ensure optimal clinical efficacy and minimal toxicity for the patient.

Clinicians and academics generally have a good understanding of the contribution of these pharmacokinetic parameters to drug dosing. However, a high-level understanding of how to adjust drug doses when there are changes to one or more of these parameters is less common. For drugs titrated to effect, e.g. vasopressor agents, dosing can be adjusted based on the observed pharmacodynamics. For other drugs where efficacy is not observable in a timely fashion, e.g. antibacterials, a ‘one dose fits all’ approach is more commonly used because of the difficulties in determining whether drug failure is occurring. Clearly, though, changes in pharmacokinetic parameters will cause a change in drug exposure during a dosing interval, thereby potentially compromising the desired drug effects.

To this end, it remains controversial as to how protein binding changes alter the time course of unbound drug concentration in plasma, and other body compartments, throughout a dosing interval. Critically, the unbound concentration is of paramount interest as it determines drug efficacy and potential drug toxicity. However, rational dose adjustment in the presence of altered protein binding is poorly understood and conflicting views exist in the literature as to the impact of changes in protein binding on drug efficacy. For this reason, clinicians often ask the following questions: (1) How important is protein binding?; (2) What proteins are involved in protein binding?; (3) Is protein binding always constant?; (4) Do pH and temperature changes alter binding?; (5) What are the clinical (practical) effects of the above?; and (6) Are protein displacement drug interactions of significance?

With this background, the purpose of this paper is to consider the data associated with altered pharmacokinetics in the presence of changes in protein binding and the clinical consequences that these may have on therapy. We use examples from the area of critical care to discuss this topic as acute changes in protein binding are common in the critically ill patient [1].

2 Unbound Fraction or Unbound Concentration?

For all drugs, it is the fraction that is not bound to plasma proteins (i.e. the unbound, or free, fraction) that is responsible for pharmacological effect. Whilst knowledge of the fraction of unbound drug is academically of interest, it is changing unbound concentrations over a dosing interval that should be described to accurately predict drug effect [2, 3]. The unbound fraction is most commonly reported as a single value to be assigned to any observed concentration for a particular drug whereas, in reality, many drugs have a different unbound fraction depending on drug concentration and various other factors [4]. It follows that data describing unbound fraction as an isolated value are in some ways not useful without a corresponding drug concentration and both should be considered together to interpret likely drug effects.

3 What is the Effect of Altered Protein Binding on Pharmacokinetics?

The unbound fraction of drug will distribute from the vascular space into tissues or be metabolised or excreted from the body. Therefore, as this fraction changes, differential pharmacokinetic effects will occur. Figure 1 schematically presents a two-compartment model for a drug, including protein binding. From this figure, it is evident that the pharmacologically active unbound concentration will be affected by the bound drug and the drug distributed into tissue, with both acting as a reservoir for unbound drug in the blood (central compartment). As drug clearance occurs, a new equilibrium between bound and distributed drug occurs, which acts to maintain the unbound drug concentration. The unbound drug concentration may decrease faster than the new equilibrium can be established if the rate of distribution from the peripheral compartment or the dissociation from protein binding is slower than drug clearance. This would be rare. The cardinal feature of the above equilibrium is that unbound concentrations are most likely to decrease later in the dosing interval where clearance has reduced the unbound drug and the reservoirs that support its concentrations. From a pharmacokinetic perspective, Fig. 1 describes how decreases in protein binding can lead to increased drug clearance as well as Vd. Therefore, for many highly bound drugs, an increased drug clearance as well as Vd is observed in states of reduced protein binding such as that commonly observed in critical illness [5].

Fig. 1
figure 1

The equilibrium between unbound, bound and distributed drug in the body in a two-compartment model. The bloodstream is the central compartment and the peripheral compartment represents the extravascular tissues where the drug distributes from the central compartment. k in corresponds to the absorption constant (in oral administration) or the infusion rate (in intravenous infusion), k 12 corresponds to the constant that describes the movement of drug from the central compartment (1) to the peripheral compartment (2). k 21 describes the movement from the peripheral compartment(s) back to the central compartment. k b and k ub describe the equilibrium between bound and unbound drug, respectively, and albumin in the bloodstream. k b and k ub will depend on the binding affinity. k b′ and k ub′ describe the equilibrium between bound and unbound drug and albumin in the peripheral compartment where binding can occur to extravasated albumin or to cell membranes (including intracellular distribution). The albumin binding equilibrium will displace depending on the plasma albumin concentration and the plasma drug concentration. k out corresponds to the elimination constant from the central compartment. Adapted from Ulldemolins et al [5], with permission from Springer International Publishing AG (© 2011. All rights reserved)

Full size image

If we further explore these concepts using pharmacokinetic equations, we can see confirmation that plasma protein binding directly affects both Vd and drug clearance [6]. Vd can be calculated using Eq. 1:

$$ {\text{V}}_{\text{d}} = \left( {\frac{{{\text{f}}_{\text{u}} }}{{{\text{f}}_{\text{uT}} }}} \right){\text{V}}_{\text{T}} + {\text{V}}_{\text{p}} $$
(1)

where fu is the fraction unbound in plasma, fuT is the fraction unbound in tissues, VT is the volume of tissue and VP is the volume of plasma. From this equation it is evident that the larger the fu, the larger the Vd, meaning that acute or chronic changes in protein binding will lead to increases in Vd. This is especially relevant for hydrophilic antibacterial agents, namely β-lactams, glycopeptides, aminoglycosides and lipopeptides, whose distribution is limited to the extracellular space, so that the amount of plasma protein binding is the major factor influencing their Vd, which is linearly correlated with the unbound fraction [7].

For clearance, drug removal is governed by blood flow of the eliminating organ (Q), fu and intrinsic clearance (CLint), which may vary in the presence of enzyme or renal tubular secretory activity. Clearance (CL) can be calculated using Eq. 2:

$$ {\text{CL = }}\frac{{{\text{Q(f}}_{\text{u}} \cdot {\text{CL}}_{\text{int}} )}}{{{\text{Q + (f}}_{\text{u}} \cdot {\text{CL}}_{\text{int}} )}} $$
(2)

For a drug cleared predominantly by glomerular filtration, clearance will increase significantly with an increase in renal blood flow because of a typically high CLint and/or with an increase in the unbound fraction. The influence of unbound concentration on clearance was well demonstrated in a group of critically ill patients treated with the highly protein bound antibacterial teicoplanin, with a significant inverse relationship between drug clearance and albumin concentrations observed [8].

From Eqs. 1 and 2, it is evident that a decrease in protein binding will lead to an increased Vd and increased clearance for drugs with high CLint. Data on individual drugs may be required to determine the magnitude of any changes in pharmacokinetic parameters and whether these translate to the need for altered dosing regimens.

4 Does the Unbound Fraction Change in the Same Dosing Interval at Different Antibacterial Concentrations?

It is unlikely for a drug to have the same unbound fraction (or unbound concentration) at all times because after administration, particularly bolus parenteral administration, the very high concentrations in plasma in this initial phase will result in disproportionately higher unbound concentrations until binding to plasma proteins can occur. This is illustrated in Fig. 2, where the unbound fraction of cefazolin is shown throughout a dosing interval [9]. In this example, the unbound fraction is highest early in the dosing interval and then as the process of drug–protein binding equilibration occurs (as well as other pharmacokinetic processes such as distribution), the unbound fraction and unbound concentration stabilizes. The clinical consequences of this may be negligible depending on the distribution and clearance of the drug as well as the changes to the unbound concentration relative to the concentrations required for therapeutic effect. Similar data have previously been shown for ceftriaxone [10] and carbamazepine [11] amongst other drugs.

Fig. 2
figure 2

Change in the fraction of unbound cefazolin in plasma during a single dosing interval (1,000 mg 30-min intravenous infusion) in critically ill patients (n = 11) with traumatic soft tissue injuries [9]. The figure shows that a higher unbound fraction is present in the first phase following drug administration. The grey squares are the observed unbound fraction of cefazolin for each patient at each timepoint and the solid line is the line of best fit for the two-phase decay in unbound fraction

Full size image

5 What Causes Changes in Protein Binding?

There are many patient presentations that will affect protein binding (Fig. 3). The likelihood of altered protein binding is more common in some patient populations such as burn injury patients, cancer, diabetes mellitus, liver and renal disease. Whether a drug will be affected in these cases depends on whether it is acidic or basic and whether it binds to albumin (typically acidic drugs) or an acute-phase reactant protein such as α1-acid glycoprotein (typically basic drugs). Some patient populations have been less well studied, although a similar prevalence of altered albumin concentrations is apparent. For instance, the SAFE (Saline versus Albumin Fluid Evaluation) study defined hypoalbuminaemia as an albumin concentration less than 25 g/L and reported that this was present in 40–50 % of critically ill patients in this study [12]. Likewise, a very high incidence of hypoalbuminaemia was observed in 200 critically ill patients with cancer, 45 % of whom had plasma albumin concentrations less than 20 g/L [1]. This suggests that altered protein binding is likely to be common for highly protein bound drugs in these acutely ill patients. Data from the antibacterials ceftriaxone, teicoplanin, cefazolin and flucloxacillin confirm these problems [5, 13–18] and we provide some example data in Fig. 4. Suffice to say, altered protein binding may occur in any of the patient groups described in Fig. 3, although the extent and clinical consequences of this may not necessarily be significant.

Fig. 3
figure 3

Main factors responsible for alterations in drug–albumin binding [52]. SIRS systemic inflammatory response syndrome

Full size image
Fig. 4
figure 4

The effect of biological variability of the fraction unbound in plasma in critically ill patients for phenytoin [n = 53] (a) and flucloxacillin [n = 10] (b). The grey squares are the observed values, the solid line represents the line of non-linear regression and the dashed lines represent the 95 % confidence interval of the non-linear regression. For both drugs, the unbound fraction increases with reducing plasma albumin concentrations

Full size image

Temperature and pH are also reported to affect protein binding. It is important to note that changes mediated by either of these environmental conditions are unlikely to be sufficient in clinical treatment to cause a change in the unbound concentration that is likely to affect the success of therapy. Some variation is expected to be seen in protein binding, but it would not significantly exceed typical inter-subject variability [19–21]. This area has been reviewed in detail by Hinderling and Hartmann [20] who hypothesized few likely clinical sequelae from pH changes, with clinically relevant protein binding from pH changes only proposed for fentanyl and lidocaine. The available data do emphasize the importance of replicating in vivo temperature and pH conditions in an in vitro setting if reliable prediction of in vivo effects of changes in protein binding is to be achieved [22].

The clinical relevance of drug displacement interactions has also been the source of controversy. In this context, it is considered that the presence of a newly introduced drug will displace from protein binding sites a drug already present in the system. What is certain is that these drug displacement interactions can occur. The consequence of any changes in unbound concentration should be interpreted as for protein binding changes caused by changes in protein concentration, pH or temperature. It follows that a drug displacement interaction will be far more likely to be problematic for therapy with a high CLint than for a drug with a low CLint. That is, clinical relevance will be determined by the pharmacokinetics/pharmacodynamics of the drug as described in Sect. 6 and 7.

6 When are Changes in Protein Binding Likely to be Therapeutically Relevant?

In the authors’ opinion, the three clinical presentations that should all be present in a patient to cause changes in the protein binding of a drug and lead to adverse clinical consequences are:

  1. a)

    highly protein bound drugs (e.g. ceftriaxone, phenytoin);

  2. b)

    high clearance drugs, particularly drugs predominantly cleared by glomerular filtration; and

  3. c)

    drugs where dosing is not titrated to effect (e.g. antibacterials).

The number of examples of drugs where pharmacokinetic changes mediated by altered protein binding are of clinical relevance is small in the context of clinical practice [23–25]. However, examples exist, particularly in acute settings, where each of the criteria discussed in Sect. 6.1–6.3 are satisfied.

6.1 Highly Protein Bound Drugs

For highly protein bound drugs, changes in the fraction bound will have a much larger overall effect on the unbound concentration. For instance, if Drug A experiences a protein binding change from 99 to 98 %, the free concentration will double, enabling more rapid clearance and extensive distribution. If Drug A changes from 99 to 95 %, then the free concentration would be assumed to increase fivefold (although the distributive processes may make this difficult to measure). However, for a drug with minimal protein binding, Drug B, a change from 5 to 4 % protein binding or even 5 to 1 % will barely affect the unbound concentration, let alone clearance and distribution pharmacokinetics. For this reason, changes in the unbound fraction of highly protein bound drugs appear to result in more significant changes in pharmacokinetics [5, 13–16].

6.2 High Clearance Drugs

It is widely agreed that changes in protein binding are unlikely to be important for drugs with low extraction ratios, i.e. those drugs where changes in organ perfusion do not significantly affect drug clearance or metabolism (e.g. carbamazepine, diazepam, warfarin). This was shown in Eq. 2, where drugs with low CLint are less likely to be affected by changes in fu. Other drugs with high CLint may result in clinically relevant increases in clearance; this is likely to be particularly problematic for drugs significantly cleared renally. Using the critically ill as an example population, Table 1 summarizes the studies that report antibacterial clearance in critically ill or healthy subjects for various highly protein bound antibacterials [8, 10, 13, 14, 16, 26–35]. Antibacterials appear to be at particular risk of significant pharmacokinetic changes. This is probably because they often have multiple half-lives within a single dosing interval and therefore clearance changes in this setting will amplify changes in pharmacokinetic exposure. Evidently, drug clearance and Vd can be particularly elevated in these patients receiving drugs with a high CLint, suggesting that the time to re-dose would be expected to be shorter.

Table 1 Changes in drug clearance for moderate to highly bound antibacterials in critically ill patients with hypoalbuminaemia compared with healthy volunteer data
Full size table

6.3 Drugs where Dosing is Not Titrated to Effect

For drugs where dosing is titrated to effect, altered exposures from protein binding changes are unlikely to be problematic, unless there is a low therapeutic window whereby small concentration increases may result in undesired supra-therapeutic effects as seen with warfarin [36]. In the case of vasopressor therapy, a change in the pharmacokinetics of an agent will be catered for by using dosing based on blood pressure, which can be measured in real time. Whilst initial dosing may be standardized for these patients, hour-to-hour dosing is based on observed blood pressure with doses adjusted to meet desired targets. For other drugs such as antibacterials and anticonvulsants where pharmacological effect is less measurable, changes in pharmacokinetics may not be identifiable unless drug concentrations are measured using therapeutic drug monitoring (TDM). For many of these drugs, of course, TDM may not be widely available, so failures of therapy may be due to decreased exposure that was not measured, resulting in a decreased effect.

To illustrate this challenge we use a recent case from our hospital in which a critically ill adult was administered ertapenem (protein binding ~90 %) 1 g intravenously daily for a ventilator-associated pneumonia caused by Enterobacter cloacae (minimum inhibitory concentration [MIC] 0.5–1.0 mg/L). The patient had a plasma albumin concentration <15 g/L and a measured urinary creatinine clearance of 220 mL/min. Using the β-lactam TDM assay available through our laboratory [37], the initial unbound trough concentration was <0.1 mg/L (our target unbound concentration was 1.0 mg/L). Eventually with incremental dosing, it was found that a dose of 1 g 8-hourly as a 4-h infusion was required to achieve an unbound concentration of 1.1 mg/L. The reason this patient had such low ertapenem concentrations was the presence of severe hypoalbuminaemia resulting in an increased unbound concentration that was rapidly cleared by supranormal renal function, otherwise known as augmented renal clearance [38–41]. The patient did not sufficiently improve with ertapenem therapy and did not demonstrate an adequate resolution of infection signs and symptoms. An alternative antibacterial was subsequently prescribed, but without having measured the concentrations we would not have known that the pharmacokinetics were so profoundly affected as no clinical symptom for infection changes sufficiently dynamically to allow appropriate dosing based on response [42]. A similar case was observed a couple of years ago in a hypoalbuminaemic transplant patient treated with teicoplanin and undergoing continuous venovenous haemofiltration [17]. While these are merely two case studies, they summarize the potential problems caused by changes in drug clearance associated with highly protein bound drugs in patients with hypoproteinaemia.

7 Relevance of Altered Protein Binding to Dosing Regimens

From the discussion thus far, it is our contention that changes in protein binding may be significant, particularly for highly protein bound and high clearance drugs where dosing is not titrated to effect. In this circumstance, it is more likely that changes in protein binding will be problematic for patients where drug concentrations are close to the efficacy or toxicity borderline.

To rationally adjust drug doses taking into account pharmacokinetic changes mediated by altered protein binding, knowledge of the pharmacokinetics/pharmacodynamics of the drug is important. To describe this concept further, we use antibacterial dosing in critically ill patients as an example. The pharmacokinetics/pharmacodynamics of antibacterials has been reviewed in detail previously [43–47] and these data emphasize that antibacterials may have concentration-dependent bacterial killing, time-dependent bacterial killing or a combination of both described with the ratio of the area under the concentration–time curve to MIC of the pathogen ratio. For all antibacterials where a larger Vd is likely because of changes in protein binding (e.g. see Table 1), larger initial doses are suggested, particularly for the first 24–48 h [48]. Such higher doses are considered especially necessary in the critically ill given the other pathophysiological changes that occur in these patients and the associated iatrogenic interventions [48, 49]. After this initial phase, dosing should be guided by the likely antibacterial clearance. For high-clearance antibacterials more frequent dosing is suggested, particularly for time-dependent antibacterials where maintaining antibacterial concentrations above the MIC for extended periods within a dosing interval is important [50].

For all antibacterials, reduced protein binding will result in higher unbound antibacterial concentrations early in the dosing interval, which should maximize bacterial killing at that time [51]. However, the increased clearance appears to result in low unbound concentrations at the end of the dosing interval. Increased clearance will again be problematic for drugs where the goal is to maintain concentrations above a certain threshold (e.g. the MIC), for instance for time-dependent antibacterials such as β-lactams. For concentration-dependent antibacterials such as daptomycin, higher dosing, rather than more frequent dosing, would be appropriate.

8 Conclusion

A change in the protein binding of drugs appears to be common in some specific patient groups and may result in variable unbound concentrations, thereby potentially affecting pharmacological activity. However, the circumstances, drugs and patients in which this is likely to be problematic may be limited to specific patient groups such as the critically ill. Specifically, drugs with high protein binding, high CLint and where dosing is not titrated to effect are most likely to be affected. The response from clinicians regarding how to adjust dosing in the presence of the associated increased Vd and drug clearance is governed by the pharmacokinetic/pharmacodynamic characteristics of drug action. We propose that further pharmacokinetic modelling-based research is required to enable the design of robust dosing regimens for drugs affected by altered protein binding.