Discussion
Variation in protein binding of antimicrobials, including piperacillin,
is well documented [18]. Despite being a moderately-bound drug, as a
beta-lactam, piperacillin unbound concentration can fluctuate in cases
of hypoalbuminemia, therefore affecting the distribution and the
excretion of renally eliminated drugs [4, 18]. This is turn may
affect the probability of target attainment in critically ill patients.
The present analysis revealed
that unbound fraction variations did not significantly affect
piperacillin total CL, and simulated concentration-time profiles showed
no impact on target attainment, as all concentrations after 24 h
(C24h) were above the PK/PD target of 1xMIC of 16 mg/L.
However, the profile of the subject
with augmented renal clearance
(ARC) (subject 4) had a rather low C24h. ARC is becoming
more prevalent in ICU and is known to cause lower serum concentrations
of drugs that are primarily eliminated by the kidneys, such as
piperacillin [9, 19, 20]. Despite successfully achieving a target of
1xMIC, our simulations indicate that patients with ARC may not be able
to reliably reach targets of 4xMIC in cases of empirical treatment,
which is often the case for piperacillin. This target is commonly used
in ICU patients in order to prevent the growth of resistant pathogens
[21, 22]. Therefore, while no dosing adjustments seemed necessary in
our analysis, it may be in cases where clinicians require the
achievement of a more aggressive target of 4xMIC.
While we investigated the impact of unbound fraction fluctuations on a
clinical level, we also evaluated the repercussions it could have on the
validity of a popPK model. Indeed, after applying various unbound
fractions, the model remained valid for values ranging from 65% to
85%. These values are what are mainly reported in studies that
determined the unbound fraction of piperacillin in critically ill
patients [3, 4, 23]. However, these same studies have also reported
values going as high as 95%. Had we used a value of 95% for example,
this model would not have been suitable for the external dataset. This
indicates that it is important to know the unbound fraction of the
population before using a popPK model developed from unbound
concentrations of piperacillin. In our case, it is possible that the
external dataset had a higher unbound fraction than the value of 70%,
indicating that the model may not be suitable at all for this
population. This shows the precautions we must take in using theoretical
factors in every analysis we perform. Nonetheless, this value offered
the best combination in terms of bias and imprecision.
This study has some limitations. Firstly, this study is limited with
simulated concentrations. Indeed, only total concentrations were
obtained from the external validation dataset, with no knowledge of the
real unbound fraction for each subject. Additionally, simulations were
performed using a popPK model that only included CLCr as a covariate on
CL, not unbound fraction, and CLCr may not fully portray piperacillin’s
total renal CL, as piperacillin is also eliminated by tubular secretion.
To the best of our knowledge, no model currently available in the
literature integrated unbound fraction, albumin levels or markers of
tubular secretion as covariates on piperacillin CL, making it difficult
to adequately evaluate their impact., possibly due to the rarity of such
information for clinicians, especially for markers of tubular secretion,
as they may not be routinely available in ICU wards.
The present study has shown through simulations that adjustments to
piperacillin dosing regimens may not be required in cases of
hypoalbuminemia. As a moderately-bound molecule, altered binding may
have little impact on piperacillin’s PK profile. Thus, despite unbound
concentration being the main parameter for target attainment, working
with total concentration may suffice for therapeutic drug monitoring of
piperacillin. However, considering the real unbound fraction of the
dataset is necessary in the case of model evaluation, as the evaluation
may be skewed if binding levels were to be assumed in the population.
Indeed, it seems it would be best to only evaluate models that used
unbound concentrations with external datasets that have real unbound
concentrations data available. A clear answer is, however, difficult to
give in the case of a molecule such as piperacillin, as opposed to
highly bound molecules in the likes of ceftriaxone or ertapenem.
Nonetheless, our simulated scenarios offer some insight regarding the
possible impact of unbound fraction variability in critically ill
patients in hopes of further evaluating this with real clinical studies.
Studies with real unbound piperacillin concentrations and unbound
fraction values would be warranted before reaching a conclusion.