Linda B.S. Aulin(1), P.H. van der Graaf(1,2), J. van Oers(3), M.W.N. Nijsten(4), E. de Jong(5), A. Beishuizen(6), A.R.J. Girbes(7), D.W. de Lange(8), J.G.C. van Hasselt(1)
(1) Leiden Academic Centre for Drug Research, Leiden University, Leiden, the Netherlands (2) Certara QSP, Canterbury, the UK (3) Elisabeth TweeSteden Hospital, Tilburg, the Netherlands (4) University Medical Center Gronongen, Groningen, the Netherlands (5) Rode Kruis Hospital, Beverwijk, the Netherlands (6) Medical Spectrum Twente, Enschede, the Netherlands (7) VU Medical Center, Amsterdam, the Netherlands (8) University Medical Centre Utrecht, Utrecht, the Netherlands
Introduction:
The use of host response biomarkers to guide antibiotic treatment of bacterial infections is well established. Procalcitonin (PCT) represents a biomarker with favorable properties, including specificity to bacterial infections, rapid induction and short half-life compared to C-reactive protein1. Previously, PCT has been shown to allow early stopping of antibiotics without impacting clinical outcomes in sepsis patients2.
Sepsis is a condition characterized by a dysregulated immune response to infections, which leads to organ and tissue damage and is associated with a high mortality3. To this end, broad-coverage AB treatment consisting of multiple combinations is typically used. Biomarkers to quantify the effect of such combinations may be relevant for treatment optimization. We hypothesized that characterization of PCT and its kinetics may be valuable to predict clinical outcomes and as a surrogate biomarker to quantify individual treatment response.
Objectives:
We aimed to evaluate early PCT as a predictor for 28-day survival and to quantitatively characterize the kinetics of PCT to assess the effect of antibiotic combination treatments in a cohort of sepsis patients.
Methods:
Study data: Data from a previously conducted randomized controlled trial investigating PCT to guide early treatment discontinuation was used (www.clinicaltrials.gov, NCT01139489). Data for 1546 patients was available with a total of 4928 PCT values (median: 6 PCT values/patient in PCT arm). Antibiotic treatments comprised of antibiotic mono- or combination therapies, with a median of 2 antibiotics used alone or in combination per patient (range: 1-7 antibiotics/patient).
Clinical outcome analysis: A parametric proportional hazard survival model was developed to describe overall 28-day survival evaluating several probability distributions to describe the baseline hazard. A univariate analysis was conducted to identify early PCT predictors of survival.
PCT kinetic model: A dynamical mixed effect model was developed to quantitatively characterize the kinetics of PCT in septic patients during and after antibiotic treatment. Unique treatments were defined based on classification at drug class level. A random effect was used to model inter-treatment variability (ITV).
Combination treatment analysis: ITVs were used to quantify individual treatment response. ITVs were analyzed using linear regression with regression coefficients for individual antibiotics present in the combination, assuming additivity. The resulting regression coefficients were used to compute pairwise expected antibiotic treatment effects. Residuals between predicted and observed treatment effect were computed in order to identify deviations from additivity.
Results:
Clinical outcome analysis: A Gompertz distribution best described overall 28-day survival. PCT was found to be a significant predictor of 28-day survival, with the absolute PCT value at day 2 being the earliest predictor (β: 3.7 x 10-3, range: 0-400 μg/L).
PCT kinetic model: A structural model with a first-order growth or degradation of PCT (kPCT) was used, and further included terms for the baseline PCT (PCT0), a delay term (PCTdelay), and a random effect term (ITVPCT) to quantify treatment associated changes in PCT kinetics within a patient. Parameter estimates were: PCT0 1.52 μg/L (RSE 9%, IIV 218.6%), kPCT 0.278 day‑1 (RSE 6%), PCTdelay 1.85 (RSE 6%), and ITVPCT estimated at a variance of 1.34 (RSE 26%).
Combination treatment analysis: antibiotic-specific treatment response regression coefficients ranged between -0.755 and 0.352 indicating associations of both positive and negative PCT kinetics. For several pairs we observed trends suggesting synergy or antagonism.
Conclusions:
We quantitatively characterized PCT and its kinetics in a large cohort of sepsis patients using time-to-event and mixed-effect modelling to obtain insight into antibiotic treatment response. This approach could potentially be generalized to other bacterial infections and host biomarkers.
References:
[1] Nijsten, M. W. et al. Procalcitonin behaves as a fast responding acute phase protein in vivo and in vitro. Crit. Care Med. 28, 458–461 (2000).
[2] de Jong, E. et al. Efficacy and safety of procalcitonin guidance in reducing the duration of antibiotic treatment in critically ill patients: A randomised, controlled, open-label trial. Lancet Infect. Dis. 16, 819–827 (2016).
[3] Rhodes, A. et al. Surviving Sepsis Campaign. Crit. Care Med. 45, 486–552 (2017).
Reference: PAGE 27 (2018) Abstr 8622 [www.page-meeting.org/?abstract=8622]
Poster: Drug/Disease Modelling - Infection