This retrospective study was conducted at the IRCCS, Azienda Ospedaliero-Universitaria di Bologna, Italy, during the period January 2021–August 2023, among patients having documented or suspected staphylococcal osteoarticular infections treated with dalbavancin monotherapy after completing an initial 2-week in-hospital daptomycin-based combination therapy. The study was approved by the local Ethics Committee (registration number 897/2021/Oss/AOUBo), and, according to hospital policies, signed informed consent was waived due to the retrospective observational design of the investigation.

All patients had an initial dalbavancin dosing regimen based on two 1500 mg doses 1 week apart (on days 1 and 8), and underwent TDM after at least 21 days from the start of treatment, as suggested in a previous study [5]. Briefly, the first TDM assessment should be performed in the time interval between 21 and 35 days, depending on the patient’s renal function, in order to be sure of maintaining, over time, dalbavancin plasma concentrations above the efficacy threshold. Dalbavancin total plasma concentrations were measured by means of a validated liquid chromatography-tandem mass spectrometry method as previously described [17]. The intra- and interday coefficients of variation of quality controls were 9.0–14.0% and 4.8–14.2%, respectively. The lower limit of quantification was 0.5 mg/L. A persisting total dalbavancin plasma concentration ≥ 8.04 mg/L was considered adequate for efficacy, as it may grant an fAUC24/MIC ratio > 111.1 against staphylococci, with an MIC value up to the EUCAST clinical breakpoint, as previously described [15]. TDM reassessments were subsequently performed whenever feasible, and additional TDM-guided dalbavancin doses were administered whenever clinically needed on a case-by-case basis. C-RP levels were assessed at baseline, i.e. at the first dalbavancin administration and at each TDM instance.

Model-informed precision dosing (MIPD) was used by means of a TDM-based Bayesian method (developed with MwPharm++ software – Mediware®) for forecasting the duration of exposure ≥ 8.04 mg/L and the eventual need for re-dosing, as previously described [18]. Clinical pharmacological advice on when to administer the next dose was then provided to clinicians. TDM-based MIPD was performed for all TDM samples, as long as a patient was treated with dalbavancin. In MIPD, TDM data along with demographic and clinical information (age, height, weight, sex, estimated glomerular filtration rate [eGFR]) were used to calculate individual dalbavancin PK parameters. The a posteriori approach was used for estimating the duration of effective exposure and for anticipating the best timing for eventual re-dosing [18].

This approach is particularly helpful for patients requiring long-term treatments and in the extremes of renal function. In fact, even if dalbavancin dose should be reduced in patients with impaired renal function as per the summary of product characteristics, when more than two doses are needed there are no specific indications on how to modify drug dose according to varying renal function.

Only those patients fulfilling the following inclusion criteria were selected for analysis: diagnosis of bone and joint infections, previous antimicrobial treatment with a daptomycin-based combination regimen, availability of at least two C-RP concentrations, of which one was available at the start of dalbavancin treatment for assessing the C-RP baseline value.

Patient data retrieved from clinical records included demographics (age, weight, height, sex), clinical data (type and site of infection, previous antibiotic regimens) and laboratory data (serum creatinine and C-RP). eGFR was calculated by means of three formulas, namely the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) [19], Cockcroft–Gault [20] and Modification of Diet in Renal Disease (MDRD) [21] formulas. The formula having the best performance in estimating dalbavancin clearance (CL) was chosen. eGFR was normalized to 1.73 m2, in agreement with what we have just done in our previous dalbavancin population PK model [5], or used without body surface area indexation.

The overall number of administered dalbavancin doses was established by the treating physician based on a test of cure (TOC), as previously described [22]. Briefly, TOC was assessed monthly by means of an ambulatory visit and was defined as positive whenever local (rubor, tumor, calor, dolor) and systemic (fever and pain) signs of infection disappeared, CRP values became normal, and there were no suggestive findings of infection on imaging studies. The dalbavancin treatment period was defined as the elapsed time between starting dalbavancin treatment and the date of positive TOC [22].

A PK/PD model was adopted to simultaneously fit both PK and PD data. Model structure is shown in electronic supplementary material (ESM) Fig. S1. The non-linear mixed-effects modeling using the stochastic approximation maximization (SAEM) algorithm implemented within the Monolix software (version 2023R1; Lixofit, Antony, France) was used for the analysis.

The structural PK model was based on a previous population PK model from our group including 69 patients having different types of subacute or chronic infections and 289 drug concentrations [5]. That study showed that the best model describing dalbavancin concentrations was the two-compartment linear model with first-order rate transfer constants between the central and peripheral compartments, and first-order rate constant of elimination from the central compartment [5]. The PK estimates of that model were used as initial values for the current model, and all the population PK parameters were re-estimated. An initial PK/PD model without covariates was built and fitted to the data. The following clinical covariates were then tested on the PK parameters: sex, weight, height, serum creatinine, and eGFR. C-RP was also tested as a continuous covariate on dalbavancin CL. Finally, a full PK/PD model that included the significantly retained covariates was built and used as the final model. The relationship between covariates and PK parameters was tested according to an exponential relationship.

PD modeling was performed using an indirect turnover model with full inhibition of C-RP production as follows (Eq. 1):

$$ \frac}R}}}t}} = kin \times \left( }} \right) - kout \times R, $$

(1)

where R represents the response (i.e., C-RP concentration in plasma); \(\fracR}t}\) represents the changing rate of C-RP concentration in plasma relative to time; Cp is the dalbavancin total plasma concentration; kin and kout represent the increasing and decreasing elimination rates of C-RP concentration in plasma, respectively; and IC50 is the dalbavancin concentration causing the half-maximal rate of C-RP decline. C-RP values at time zero (R0), i.e. when starting dalbavancin treatment, were considered as the baseline values for modeling As C-RP concentration at baseline is independent from dalbavancin concentration, R0 was fixed to the value of C-RP for the ith individual at time zero, therefore only the inhibition production of C-RP is estimated. R0 is equivalent to the kin/kout ratio (R0 = \(kin/kout\)). The choice of this type of model was based on the biological plausibility and on previous population PK/PD studies that investigated the relationship between vancomycin or teicoplanin concentrations and C-RP [23, 24].

All individual parameters were considered to be log-normally distributed, random effects were normally distributed, and an exponential model was used for describing the individual parameter estimates. Different error models (constant, proportional, or combined error model) were tested for describing the residual variability.

Covariate analysis was conducted according to a forward/backward process. In the forward step, the correlation between each covariate and the random effect of the estimated PK parameter was tested. In particular, Pearson’s correlation test was used to test the null hypothesis that the correlation coefficient between the random effects (calculated from the individual parameters sampled from the conditional distribution) and the covariate values is zero. In the backward step, each covariate is tested for removal from the full covariate model based on the Wald test. Both tests were considered positive when the p-value was < 0.05. Finally, a covariate was included in the final model if a decrease of ≥ 3.84 points in the objective function value (OFV), coupled with a reduction in the Akaike information criteria (AIC) and Bayesian information criteria (BIC), were observed with respect to the base model. Of note, these two latter criteria are indicated in the case of models including covariates for adjusting for model complexity.

Reliability of the PK/PD model was based on the observation of < 30–40% relative standard errors (RSE%) of the PK and PD estimates, and on the adequacy of the following goodness-of-fit plots: observation versus population and individual predictions, usual residual-based plots (individual- and population-weighted residuals), and prediction-corrected visual predictive check (pcVPC).

Overall, 1000 non-parametric bootstrap iterations with resampling for each estimated PK and PD parameter were simulated using the Rsmlx package of R (R speaks Monolix), and the median (interquartile range) were reported. The observed versus predicted concentration for both the PK/PD and the pcVPC were replotted in R.

First, the relationship between dalbavancin concentrations and the extent of C-RP production inhibition was assessed by means of 10,000 subject Monte Carlo simulations based on the population variability and between-subject variability resulting from the PD model. The median (5th–95th percentiles) probability of inhibiting CRP production was calculated in the presence of dalbavancin concentrations ranging from 0.5 to 100 mg/L, with 0.5 mg/L being the lowest quantifiable dalbavancin concentration.

Second, for each of the tested dosing regimens, the median of the percentage reduction extent of the C-RP value at each specific date versus baseline was assessed. The probability of achieving definitive C-RP production inhibition (defined as values < 1 mg/dL) was also assessed.

Three different dosing regimens of a cumulative dalbavancin dose of 3000 mg over 3 weeks, according to what was proposed by Senneville et al. [16] (namely 1500 mg on day 1 + 1500 mg on day 8; 1500 mg on day 1 + 1500 mg on day 15; 1500 mg on day 1 + 500 mg on day 8 + 500 mg on day 15 + 500 mg on day 22) were tested in relation to four different classes of renal function (eGFR of 0–29, 30–59, 60–89, and 90–120 mL/min). An additional dose of 1500 mg administered on day 43 was simulated to test the same objectives in the case of patients clinically needing an extension of treatment duration over 6 weeks.

Monte Carlo simulations based on the final PK/PD model were performed by means of Simulix 2023R1. A total of 10,000 C-RP concentration versus time profiles were generated for each dalbavancin dosing regimen in relation to the different classes of eGFR. The estimated PK/PD population parameter values with interindividual variability (omega values) were considered for simulations. The same seed of reproducibility was used for all simulations.