The main results of the present study are: 1) our protocol seems to consent to a fast achievement of a target VSC in patients with normal kidney function and in those with kidney dysfunction; 2) the presence of an augmented renal clearance was the main determinant of the difficulties in reaching a target VSC.
Early antibiotic administration should aim to reach adequate target VSC within a few hours from infusion [10, 17, 19]. However, critically ill patients usually have a large volume of distribution which likely reflects significant capillary leakage. The latter, coupled with aggressive fluid loading, can expand the interstitial space [4, 20]. Hence hydrophilic antimicrobials — distributed exclusively in the extracellular compartment— are expected to be diluted. This shift of fluid may favour movement of drug into the interstitium and a decreased VSC is expected. Moreover, while critical illness might alter the volume of distribution, renal dysfunction additionally makes antibiotic pharmacokinetics even more unpredictable . As a consequence, patients with decreased renal function require vancomycin to be administered following dose adjustments. Hence our protocol was designed in order to avoid both a VSC under and over the target range, although we choose to favour the avoidance of a VSC under the range since its association with an increased in-hospital mortality. At the same time we felt that patients with kidney dysfunction should have had a specific protocol because of reduced renal clearance. Finally, target VSC may be difficult to achieve because of the presence of augmented renal clearance [9, 12, 21]. ARC is characterised by an enhanced renal elimination of circulating solutes observed in critically ill patients [8, 22]. The presence of ARC might imply subtherapeutic levels of a given drug for substantial periods of the dosing interval resulting in treatment failure or selection of resistant organisms [1, 12, 22].
All the above mentioned factors imply great inter-individual variability in pharmacokinetics, complicating accurate prediction of serum concentrations in ICU patients, making evident the need of a frequent dosage of VSC in order to make the daily dose of this antibiotics adequate .
A below-range VSC was observed in only 20 % of patients of both groups. However, only patients with kidney dysfunction progressively and significantly reduced this percentage, which became about 9 % at the third VSC determination. Interestingly, the group with normal kidney function had a different behaviour. Between the first and the third VSC determination, it became progressively evident the relevance of an ARC that did not allow this group to decrease the percentage of patients with VSC under the range. This is of clinical relevance since patients who reached a subtherapeutic level at the first VSC measurement had a significant correlation with in-hospital mortality (OR 2.1; p 0.003).
As to the VSC in the desired range, only at the first determination group B had higher percentage of patients with VSC in the normal range and, as a consequence, a lower percentage with a VSC over the range (Fig. 2). At the second and the third VSC determination the two groups were almost identical. This was the case also for VSC over the target range, being the percentage of patients with a VSC higher than 30 mg/L of about 28 %. The increase of VSC over the target range has been associated with a risk of nephrotoxicity that has a reported incidence up to 35 % during vancomycin therapy [23, 24]. Hence we investigated the effects of VSC over the target range on renal function. Interestingly no correlation was found between VSC and renal toxicity, even in patients with kidney dysfunction. This is relevant from the clinical point of view, since other and more expensive antibiotics are usually proposed in patients at risk of kidney dysfunction. Such result can be explained by several factors. First of all, we considered 15–25 mg/L as a target range since a VSC over 25 mg/L can be associated with increase of nephrotoxicity, as previously described . However, other studies have proposed an upper limit of 30 mg/L , suggesting that nephrotoxicity could be enhanced when the VSC reaches values higher than 30 mg/L, as it was the case of the present study. The VSC > 30 mg/L was found in about 30 % of the patients of both groups (Fig. 2), whilst the % of patients with a VSC higher than 35 mg/dl was extremely low in both groups (about 1 %, see results section). These results underline that our algorithm is able to adjust the VSC preventing the progressive accumulation of vancomycin without incurring the opposite phenomenon, that is a VSC below 15 mg/dl.
Our algorithm is based on a continuous infusion of vancomycin that might have advantages over intermittent administration [25, 26] since this strategy appears to reduce renal toxicity . Despite previous studies comparing continuous and intermittent administration have reached conflicting results , others demonstrated that continuous infusion of vancomycin is less expensive and quicker in achieving target concentration, resulting in less variability in serum concentrations [27, 19].
The presence of patients with a VSC > 30 mg/L implies that our algorithm should not be changed by increasing the daily dose of vancomycin in order to avoid a VSC under the desired range. This is suggested by the AUC/MIC ratio, which was even at the first VSC determination much higher than 400 (Table 2). This is of clinical relevance since Holmes et al  have previously demonstrated a 12 % lower mortality 30-day mortality in patients achieving a vancomycin AUC/MIC of >373 within the first 96 h of vancomycin therapy compared to those who did not. Instead, we believe that our algorithm should take into account the presence of ARC, therefor increasing the dose only in patients that really need this adjustment of the therapy.
Limitation of the study
We used a loading dose of about 15 ml/Kg of actual body weight to avoid too high VSC. However, the LD varies among different studies and other authors suggest a LD of about 25–30 mg/Kg (low level of evidence – III - and grade of recommendation B) , with the aim of rapid achievement of the target VSC. Looking at our data, between 40 and 50 % of patients of both groups were over-range, generating a clinical dilemma. Should we modify our algorithm by increasing the loading dose or should we maintain the LD used in the present study? Indeed, it can be hypothesized that higher LD would have determined higher plasmatic concentration. Hence it could be expected an increased % of patient in over-range or even of patients in over range with VSC much higher than those obtained in the present study, leading to an increase of nephrotoxicity. Moreover, the AUC/MIC was much higher than 400 for both groups Table 2), suggesting that the LD might have been sufficient to reach the expected VSC. Further studies are required to clarify if an increased LD could decrease the % of patients in under-range without increasing nephrotoxicity.
Finally, calculation of creatinine clearance implies determination of urinary creatinine. The latter, however is influenced by the volume status of the patient, treatment with the loops diuretics and vasopressor agents, and release of antidiuretic hormone. Indeed, correct estimation of glomerular filtration rates implies its determination by using inulin or iohexol clearance, and radionucleotide. Unfortunately, these methods are largely unavailable in the clinical setting.