Anti infective drugs during continuous hemodialysis – using the bench to learn what to do at the bedside

Methods: Ceftazidime, ciprofloxacin, flucloxacillin, gentamicin, linezolid, meropenem, metronidazole, piperacillin, rifampicin, vancomycin and voriconazole were studied in two different

All content following this page was uploaded by Alexander Brinkmann on 02 August 2015.

Anti-Infective Drugs during Continuous Haemodialysis – Using the Bench to Learn What to Do At the Bedside

Background:

Significant uncertainty remains for anti-infective dosing regimens in critically ill patients undergoing renal replacement therapy (RRT) The main objective of this study is to investigate the clearance of eleven selected anti-infectives in sodium chloride 0.9% and in human serum albumin (HSA) 5% in an

in vitro model of continuous venovenous haemodialysis (CVVHD), in order to suggest rational dosing strategies for clinical practice

Methods:

Ceftazidime, ciprofloxacin, flucloxacillin, gentamicin, linezolid, meropenem, metronidazole,

piperacillin, rifampicin, vancomycin and voriconazole were studied in two different solvents (sodium chloride 0.9% and HSA 5%) using a Multifiltrate® dialysis device by Fresenius Medical Care (mode: CVVHD; blood flow 100 ml/min, dialysate flow 2000 ml/h, no ultrafiltrate, filter unit: Ultraflux AV 600S) For each solution, prefilter, postfilter and dialysate samples were drawn simultaneously during one hour of dialysis Drug assay was performed with validated HPLC with UV-detection and FPIA

Results:

The clearance of ceftazidime, ciprofloxacin, flucloxacillin, gentamicin, linezolid, meropenem,

metronidazole, piperacillin, vancomycin and voriconazole in sodium chloride 0.9% was comparable (mean 1.76 ± 0.11 l/h) The clearance of these agents in human serum albumin solution 5% was reduced by between 5.3 and 72.2% The unbound drug fraction correlated with a lower clearance in HSA 5% (Pearson correlation coefficient r = 0.933; p = 0.00008) No correlation between clearance in HSA 5% and the drugs’ molecular weight was found (Pearson correlation coefficient r = 0.388; p = 0.268) Rifampicin was detected to bind to the surface of the polysulfone filter used (98% adsorption within 15 minutes) Dialysis clearance of ceftazidime, gentamicin, linezolid, meropenem,

metronidazole, piperacillin and vancomycin during CVVHD accounted for over 25% of the total body clearance ofpopulation pharmacokinetic data for renally impaired patients

Conclusion:

In vitro studies are useful to predict likely in vivo drug clearances The results from this study

highlight that dose adaptations are needed for most of the drugs under investigation for patients undergoing CVVHD, in order to optimise anti-infective dosing regimens Rifampicin should be avoided in patients undergoing CVVHD with polysulfone filters, or should only be used under the guidance of therapeutic drug monitoring to ensure adequate serum levels

Keywords: antimicrobial agents, dialysability, drug dosing, in vitro, pharmacokinetics, renal

replacement therapy

Short summary: This study uses an in vitro model of continuous haemodialysis to describe the main

factors affecting the elimination of anti-infective drugs We found that dialysis clearance depends primarily on protein binding rather than on the drugs’ molecular weight We also found that rifampicin binds to the surface of the polysulfone filter under investigation

Introduction

Anti-infective dosing regimens for critically ill patients undergoing renal replacement therapy (RRT) remain an area of uncertainty in intensive care medicine Although the timely (1) and appropriate (2) administration of antibiotics is known to save patients’ lives, only few rational dosing guidelines on how to optimise treatment during RRT are currently available to clinicians Approximately 5% of critically ill patients undergo continuous RRT (3), with this proportion increasing steadily over time

In the past decade, the advent of new dialysis techniques has improved the tolerability of RRT, whilst also changing the efficiency with which solutes like drugs are removed A large number of clinical pharmacokinetic studies have been developing general anti-infective dosing recommendations based

on rather small numbers of patient, accepting that this particular patient population is subject to huge pharmacokinetic variability (4) The conflicting results on drug clearances presented by these studies (5) pose a real challenge to clinicians Therefore, controlled in vitro studies are considered to be an ideal way of characterising RRT-drug interactions

Furthermore, the pharmacokinetics in patients with acute or chronic kidney diseases undergoing dialysis are commonly altered (6) due to physiological changes Especially, decreased protein binding may affect drug clearances and, as a result, the serum levels of agents characterised by high plasma protein binding The enhanced drug clearance resulting thereof carry the risk of producing undesirably low serum levels of anti-infective agents (7)

The aim of this study was to investigate the dialysis clearance (CLHD) of eleven anti-infective drugs during continuous venovenous haemodialysis (CVVHD) in order to assess correlations between drug clearances and protein binding (PB) and the drugs’ molecular weight We also aim to use these results

to suggest dosing regimens for these drugs in common CVVHD settings

Subjects and Methods:

In vitro model of haemodialysis:

1000 ml saline 0.9% solutions and 500 ml human serum albumin 5% (HSA) solutions served as reservoirs for different anti-infective drugs in the RRT model The baseline concentrations were as follows: ceftazidime 80 mg/l, ciprofloxacin 15 mg/l, flucloxacillin 80 mg/l, gentamicin 20 mg/l, linezolid 20 mg/l, meropenem 20 mg/l, metronidazole 30 mg/l, piperacillin 80 mg/l, rifampicin

10 mg/l, vancomycin 40 mg/l and voriconazole 20 mg/l Initial concentrations matched high standard levels, or even higher levels, to ensure appropriate chromatographic conditions, as an extreme loss of substances was predicted to occur during the time of dialysis

RRT was performed in line with clinical guidelines using a Multifiltrate® machine (Fresenius Medical Care) All experiments were performed with either saline or HSA solutions

Drug solutions were mixed and maintained stable at 37°C by using a magnetic stirrer The dialysis mode chosen was CVVHD, blood flow was 100 ml/min, dialysate flow was 2000 ml/h, and there was

no ultrafiltration The filter type used was a polysulfone AV 600S (1.4 m2, Fresenius Medical Care) device; no anticoagulation was applied The dialysis fluid used was bicarbonate buffered solution (MultiBic 2 mmol/l potassium by Fresenius Medical Care)

Sampling

500 µl of arterial, venous and dialysate samples were drawn simultaneously at 0, 5, 15, 25, 35, 45 and

55 minutes; the time to reach equilibrium before the start of the dialysis session was 3 minutes for all solutions

Analytical methods:

Gentamicin and vancomycin were measured using a Fluorescence Polarisation Immuno-Assay (AxSym, Abbott Laboratories) To quantify ceftazidime, ciprofloxacin, flucloxacillin, linezolid, meropenem, metronidazole, piperacillin, rifampicin, and voriconazole, HPLC methods with UV detection were used All HPLC methods have been previously described (8-11) and were in line with the Valistat 2.0 validation criteria as required by the German Society of Toxicology and Forensic Chemistry (GTFCh), linearity was proven to be within the requested concentration range across all drugs, and quality control samples at three different concentrations featured a relative standard deviation (RSD%) for inter- and intraday precision of <10% across all methods An external standard method was used to quantify drug concentrations, and no protein precipitation was performed for the aqueous solutions

The protein binding of drugs in HSA was determined by ultrafiltration Spiked HSA 5% and NaCl 0.9% solutions were centrifuged (centrifree® YM tubes by Milipore) for 30 minutes at 2,500 rpm and 25°C Ultrafiltrate, HSA and saline samples were all assayed

where ke is the linear regression gradient, t1/2 the corresponding elimination half-life, Vr the drug reservoir volume and ClHD;drug the resulting CLHD

The saturation coefficient (Sd) was calculated using the formula

with cdialysate, carterial and cvenous representing the concentrations in the dialysate, the arterial and venous samples, at the same sampling time point respectively

HSA protein binding was calculated using the equation

where cfiltrate HSA;drug and cHSA;drug represent the drug concentrations in the ultrafiltrate and in HSA, respectively

The statistical analysis was performed using Pearson’s correlation with SPSS® version 21, where a value of <0.05 was considered to be significant

P-Suggestions for anti-infective doses in patients with renal dysfunction and/or CVVHD are presented in Table 2 Population pharmacokinetic data (12) of patients without renal impairment were used for standard body clearances for a given drug If a range was provided in the pharmacokinetic data, the mean value was used Subsequently, the clearance values for renal impairment with a residual renal function of 15 ml/min were calculated (CLtot) using the Q0 concept by Dettli (13), which relates the reduction in total body clearance to the drug fraction eliminated by renal clearance The measured

CLHD in HSA 5% was then added to the reduced body clearance to reflect likely in vivo

circumstances The change in the standard doses was calculated using the following equation:

where Ddrug represents the suggested drug dose in mg, Dstandard a regular dose used in a clinical setting,

CLtot the body clearance without renal impairment and CLtot15 the body clearance with renal

impairment

Results

This study assessed 209 drug samples The gradual decreases in drug concentrations in sodium chloride solutions were in line with first order kinetics The dialysate samples ranged between post- and prefilter samples CLHD rates in NaCl 0.9% were comparable and ranged from 1,61 l/h for vancomycin to 1,92 l/h for meropenem, while CLHD rates in HSA 5% were reduced across all drugs (between 0.45 l/h for flucloxacillin and 1.76 l/h for meropenem) The results as to elimination half-life, saturation coefficient, CLHD in sodium chloride/HSA and PB in HSA are summarized in Table 1

As opposed to other drugs, rifampicin showed an extremely high loss of substance (98%) within the first 15 minutes of dialysis No substance at all was detectable in the dialysate, and the colour of the filter changed visibly from white to brown after the start of the dialysis session As this behaviour has

so far gone unreported, the data derived from rifampicin was not considered in the calculation of dialysis properties

There was a highly significant correlation between the decrease in CLHD in HSA and the

corresponding PB (Pearson correlation coefficient r = 0.933; p =0.00008, see Fig 1), with a similar correlation found for reported plasma PBs (r = 0.915; p = 0.00021, see Fig 1) However, bindings in HSA were 14.12 % (mean) lower than the upper limit of the values reported (12) for plasma PB

No correlation between the molecular weight and the clearance rates in saline solution was found (Pearson correlation coefficient r = 0.388; p = 0.268, see Fig 2)

For ceftazidime, gentamicin, linezolid, meropenem, metronidazole, piperacillin and vancomycin, the

CLHD calculated accounted for over 25% of the total body clearance in given standard patients The variation of standard doses versus the theoretical and actual plus theoretical total clearances is shown in Table 2

The experimental protocol for this in vitro study was chosen in line with the CVVHD setting most commonly used on our ICU Drugs from different groups were selected to cover a broad range of pharmacokinetic variability

For ceftazidime, ciprofloxacin, gentamicin, meropenem, metronidazole, flucloxacillin, linezolid, piperacillin, vancomycin and voriconazole, free diffusion with a non-concentration-related

transmembrane elimination was observed Elimination occurred following first order kinetics as described in previous publications (3, 14) The free diffusion pattern was confirmed by the emergence

of matching amounts of drugs in the dialysate and by the saturation coefficient of around 1 across all substances As has been suggested by theoretical considerations (15), our findings demonstrate that molecular weight has no measurable influence on CLHD in the protocol used The pore sizes of currently used filter materials like polysulfone are far larger than the size of commonly used

antimicrobial agents

The binding of drugs to different dialysis filters has already been proven In fact, it is well known that the binding of aminoglycosides to polyacrylonitrile membranes is related to their concentration (16, 17) Our finding that rifampicin binds to the surface of the polysulfone filter by a yet unknown

mechanism was neither expected, nor has it been previously described elsewhere This emphasizes the importance of in vitro models in the investigation of different filter materials, as information on one material cannot be generalised for other materials In vitro models are easy to conduct, fast and

inexpensive, on top of quickly delivering information on established and new agents

The choice of human albumin solution was due to the fact that albumin is the major carrier of drugs in the blood (18) However, as albumin is not the only endogenous compound with a drug binding capacity, a mean PB that was constantly below values reported for whole blood was seen One

example is flucloxacillin with a measured value of 66,6% vs > 90% as reported in literature This has also been described by Isla et al (19) in an in vitro model of CVVH and CVVHD, where the free fraction of ceftazidime in Ringer Lactate Solution plus bovine albumin was higher than in human plasma Still, reduced PB is a widely known phenomenon in critically ill patients (20) The CLHD rates

in HSA 5% and their correlation to the amount of PB indicate that protein binding is a key factor in drug loss during the dialysis process

Drug elimination during RRT is generally considered to be a major concern Hence, dose adaptations should be envisaged if drug elimination exceeds one fourth to one third of total body clearance (3, 21) According to the findings, ceftazidime, gentamicin, linezolid, meropenem, metronidazole, piperacillin and vancomycin could potentially be underdosed if administered at the reduced doses recommended in renal impairment, without compensating the loss of dialysate On top of that, this study shows that – contrary to the generally accepted rules - some drugs are eliminated more effectively in dialysis patients than in patients with no renal impairment This might also be the case for metronidazole with its relatively low PB and small amount of renal excretion In line with our findings, the phenomenon that metronidazole is removed more effectively, with a higher total CL, in patients under intermittent haemodialysis vs healthy subjects, has already been described in a previous study on four patients (22)

The study protocol was limited by the fixed dialysis doses used for the investigation, which might not reflect real life circumstances on the ICU Linearity between the CLHD rates and the applied dialysate flow, which has been discussed quite controversially in literature, may depend on the filter surface area and material Filter surfaces > 1.4 m2 are unlikely to impact the amount of drug dialysed, as complete diffusion has been demonstrated Smaller surface areas, on the other hand, which may no longer be used, could reduce the amount dialysed This might also apply to extremely high blood and dialysate/filtrate flow rates, as occurring under intermittent methods, due to a shorter contact time at the membrane

The non-linear pharmacokinetics of voriconazole are highly unpredictable (23) Although the amount

of drug elimination during the dialysis process is fairly small compared to total body clearance, it should be monitored by TDM, regardless of RRT The same applies to the use of rifampicin in

CVVHD patients with polysulfone filters

The main limitation of this study is the in vitro design In critically ill patients, a high variability in the pharmacokinetics of anti-infective drugs has been described (24, 25), with alterations in volume of