DISCUSSION
Our results demonstrate that the clearance of levetiracetam in critically ill patients undergoing CRRT is consistently demonstrated in the literature to be similar to that seen in healthy adult patients. Both mean total clearance (3.55 L/hr) and elimination half-life (9.41 hrs) were broadly equivalent to that of healthy adults (4.03 L/hr and 6 – 8 hrs respectively). The mean clearance attributable to CRRT was over 50% confirming that CRRT is responsible for the majority of drug clearance. Moreover, it demonstrates that IHD and CRRT have similar pharmacokinetic effects in terms of their ability to eliminate levetiracetam. Taken together, these findings suggest that current dosing recommendations may pose a risk of sub-therapeutic drug concentrations.
Our simulation data also raise several concerns over the current UK dosing recommendations in CRRT, which advise a 250 – 750 mg twice daily regimen without an initial loading dose [17] . As demonstrated in Fig. 5 , without a loading dose the majority of our simulated patients experienced sub-therapeutic drug concentrations for up (and beyond) 72 hours. This effect was reduced to a limited extent with higher dosing regimens (750 – 1000 mg) but even at higher doses, drug concentrations remained sub-therapeutic for the initial 24 hours of treatment. Even the addition of the loading dose recommended for patients on IHD resulted in the same problem, with the majority of patients experiencing sub-therapeutic drug concentrations up to 24 – 36 hours into treatment (Fig 4 ). The addition of a standard loading dose at 60mg/kg resolved this problem and resulted in the immediate achievement of therapeutic drug concentrations (Fig. 3 ). A significant proportion of patients at this loading dose, however, had trough concentrations over 80 mcg/mL for the first 24 hours and may require additional monitoring for potential drug toxicity[41] .
Regular dosing at lower levels (250 – 500mg) was unlikely to achieve therapeutic levels by steady state (Fig. 3 – 5 ). Even with the addition of a loading dose, the median drug concentration for patients receiving these doses remained below 12 mcg/mL at 72 hours. In contrast, at higher doses (750 – 1000mg) a significant number of patients had therapeutic drug concentrations at steady. However, without a standard loading dose of 60mg/kg the time to therapeutic levels (even at higher doses) was delayed by up to 36 hours.
From a pharmacokinetic perspective, these data suggest that current UK dosing recommendations in CRRT are likely to result in extended periods of sub-therapeusis and suggests that patients undergoing CRRT should be considered for a twice daily dosing of 500 – 1,000 mg in addition to an initial loading dose of 60mg/kg. A major limitation of this work is the lack of data on efficacy or the use of the target trough concentration in the critically ill cohort. However, given the relatively low event rate of seizures, the relatively low number of patients receiving the drug whilst also receiving renal replacement therapy and the number patients receiving this drug for prophylaxis, it is highly unlikely in our view that a PK-PD trial will be realised.
The specific extrinsic factors related to CRRT, such as effluent flow rate or dialysate rate, that influence levetiracetam clearance remain unclear. Few studies investigated the haemofiltration characteristics that influence clearance with the exception of Kalaria et al. Their analysis identified effluent flow rate and sieving co-efficient as the main influences on clearance. Effluent rates in excess of 3.5L/h were also associated with increased clearance and so higher doses may be required in these patients.
Similarly, further investigation into the intrinsic patient-specific factors that affect clearance is also required. While mean clearance was consistent between studies, there was a wide variation in individual patient clearance and plasma concentrations which are not easily explained by effluent rate and sieving co-efficient alone. Six patients demonstrated trough drug concentrations below the therapeutic target level but a post-hoc descriptive analysis (Appendix 3 ) found no obvious explanation. Three of these patients had significant residual urine output (>300ml per day), while three others had high effluent rates. However, other patients received similar doses at higher effluent rates or indeed had higher urine outputs without their plasma concentrations falling outside of target levels. These data suggest that there is a high inter-person variability of drug pharmacokinetics and determining optimum dosing strategies for individuals may require therapeutic drug monitoring.
Finally, our systematic review illustrates that the availability of pharmacokinetic data on levetiracetam in CRRT remains limited. Data are limited to a small number of case reports and two small prospective studies. There is a lack of consistency between studies, both in terms of methodology and data reporting, that further limits the overall evidence base at present. In this paper, we have introduced a novel quality assessment method that utilises a combination of the QoE framework, which assesses the strength of pharmacokinetic studies based on the quality of data and pharmacokinetic modelling provided, and the ADQI minimum reporting criteria, which was developed to standardise comparisons between studies that reported on CRRT techniques. Both methods have been used in similar systematic reviews in this way[42, 43] but never before in combination. By combining these assessment methods, our quality assessment method allows for equal consideration of both the pharmacokinetic and haemofiltration data. This enables an assessment of both the quality of the data itself and the ability to perform meta-analysis.
Further study is required in order to understand the high inter-person variability in levetiracetam clearance and the in vivo effects of different dosing strategies. Such data can be used to develop pharmacokinetic models that will enable clinicians to determine patient-specific dosing strategies. Future in vivo studies should prioritise the collection of standardised data to facilitate meta-analysis and modelling. Our recommendation is that this should include the ADQI minimum reporting criteria alongside the following: (1) IBW; (2) urine output; (3) baseline bloods (albumin, creatinine, haematocrit, liver function); (4) dosing strategy; (5) a sampling strategy that at a minimum includes pre-filter, post-filter and effluent levels; (6) information on whether seizures were controlled or uncontrolled. Standardisation will enable more robust modelling than is possible at present.
In addition to in vivo studies, experimental study using lab-based models that simulate critically ill patients undergoing CRRT may be valuable. These studies, such as the latest ex-vivo in-vivo study by Kalaria et al. [44] , would be useful in evaluating the effects on clearance of different CRRT parameters and levetiracetam dosing without the potential risk of exposing patients to possible sub-therapeutic or toxic drug levels.