In patients with septic shock, we demonstrated that the median volume of non-resuscitation and resuscitation fluids was similar at day 1 and that non-resuscitation fluids represented the largest source of fluid from day 2 and onwards after admission to the ICU. Daily fluid balance was positive until day 3 after admission. Fluid as a vehicle for intravenous drugs constituted the major contributor to non-resuscitation fluids.
Resuscitation fluids are administered rapidly to increase preload by an increase in intravascular volume, with crystalloids most commonly administered for this purpose. However, crystalloids may also be administered at a lower rate to provide hydration and to maintain homeostasis, commonly referred to as maintenance fluid. To separate these two indications, we defined a crystalloid as a non-resuscitation fluid if administered at a rate of less than 5 ml/kg/h. This definition aligns with several previous surveys on fluid administration practices [8, 12, 14]. However, it should be noted that some of the studies mentioned below have used other definitions. For instance, crystalloids were defined as a non-resuscitation fluid if administered for reasons other than circulatory impairment  or as a resuscitation fluid regardless of infusion rate  and in some studies classification of crystalloids was not described in detail [4, 9]. Accordingly, fraction of crystalloids of the non-resuscitation fluids may differ somewhat and highlight the need for a consensus with regard to definitions. Because crystalloids are generally a small fraction of total volume of non-resuscitation fluid, this will not influence the conclusions below.
Previous studies surveying the use of non-resuscitation fluids in patients admitted to ICUs have included a broad set of critically ill patients regardless of their hemodynamic status and need of resuscitation fluids [4, 8, 9]. In this study, we focused our efforts on patients with septic shock, because this is a subgroup of critically ill patients in which the importance of fluid resuscitation is emphasized in guidelines and is also a group where positive fluid balance tends to be a clinical problem. Very little is known about use of non-resuscitation fluids in septic shock, but some data can be extracted from a pilot trial (CLASSIC), comparing two protocols for administration of resuscitation fluids in septic shock . In that study, patients were included approximately 4 h after admission to the ICU and non-resuscitation fluids constituted the major part of the administered fluid already on day 1. Our finding show that this is true also if the early phase of ICU admission, during which resuscitation fluids are most likely to be administered, is included in the day 1 data. The daily fluid balance remained positive up to the 3rd day after admission in our study, in line with previous data from patients with septic shock . With regard to total volume of non-resuscitation fluid during the 5-day observation period, the two studies differ substantially. Patients in the present cohort received approximately 8 l and patients in CLASSIC cohort received approximately 11 l. Differences may be explained by variations in local practice but could also reflect a general change in practice patterns due to increasing awareness of potential adverse effects of intravenous fluids. Another possible explanation is that the median length of stay in the present study was 3 days compared to 6 days in the CLASSIC cohort. Other recent multicenter studies of cohorts of mixed ICU patients reported the total volume of non-resuscitation fluids during the first 3 days of admission to be 5–5.5 l which is similar to our finding of 6.0 l during the same time interval [4, 9]. Similar to our results, these studies also reported that vehicles were the largest fraction of the non-resuscitation fluids.
Based on the growing concern for the adverse effects of excessive fluid administration, several recent pilot studies have assessed if administration of resuscitation fluids can be reduced by “restrictive” protocols or by prediction of fluid responsiveness prior to administration of fluids [5,6,7, 15]. Data from these studies suggest that the volume of resuscitation fluid can be reduced by 0.8 to 1.2 L during the first 3–5 days of ICU admission. Our results, suggesting that administration of non-resuscitation fluid theoretically could be reduced by a median of 2.8 l during the first 5 days in the ICU, indicate that this approach could potentially have an even larger impact on fluid balance in the ICU.
Several aspects of our modelling could be considered. First, the choice to omit the use of maintenance fluids or glucose could be questioned. However, we are not aware of any studies supporting the use of maintenance fluids in patients that are in a positive fluid balance, nor are we aware of any studies or guidelines suggesting that intravenous glucose should be administered during the acute phase of critical illness [16, 17]. Second, our data suggested that a large part of the vehicles were administered as diluents of antibiotics and vasoactive drugs. To model the potential to reduce vehicles for antibiotics, one could consider both the potential to reduce vehicle volume in already used antibiotics as well as a shift away from antibiotics requiring large volumes of vehicle. Because of the large number of different antibiotics and very limited data on solubility and safety of concentrated antibiotic solutions, such modelling would have been very complex and we therefore elected not to do it. Similarly, we did not model potential reductions in administration of vasoactive drugs, because this group of drugs consists of several different individual drugs with different potentials for concentration. This means that we could have underestimated the potential to reduce the volume of non-resuscitation fluids. Third, the assumption that reductions in volume of non-resuscitation fluid would not be offset by increased administration of resuscitation fluids could be questioned, since crystalloid maintenance fluids will also distribute in the intravascular space. However, intravascular retention of crystalloids over time is most likely very low and is reported to be < 10% in inflammatory conditions meaning that this source of error is reasonably small [18, 19]. Lastly, the physicians caring for the included patients may have aimed for a positive cumulative fluid balance when prescribing non-resuscitation fluids in patients with a perceived preexisting fluid deficit. If so, we may have overestimated the potential to reduce the volume of non-resuscitation fluids somewhat in those patients.
The considerable variation between sites regarding the administration of the different subtypes of non-resuscitation fluids observed in the present study aligns with recently reported data from a multicenter retrospective study in centres in the UK and Canada . This indicates that local practice traditions, rather than evidence-based medicine, play a major role in determining volume as well as the type of non-resuscitation fluid, and highlight the need for more knowledge in this aspect of fluid therapy. Interestingly, none of the sites had any written guidelines with regard to intravenous administration of maintenance fluids or glucose solutions in septic shock. The high variability of current practice has its implications on the design of interventional trials, due to the difficulties in defining a common baseline to which an intervention can be compared. Nevertheless, we believe that our results provide a rationale for an interventional study in which a more restrictive approach of administration of non-resuscitation fluids can be compared to current practice.
While ongoing trials are addressing if restricting resuscitation fluids in septic shock impacts survival [20, 21], an important aspect to bear in mind is that the balance between benefit and harm when reducing resuscitation fluids may be different than the balance when reducing non-resuscitation fluids. Thus, it is important that any intervention with the objective to reduce administration of non-resuscitation fluids should be rigorously assessed in trials regardless of the findings in the ongoing resuscitation fluid trials.
Strengths and limitations
Strengths of our study include that data are contemporary and collected from multiple sites in both university and regional hospitals. Moreover, the study was also prospectively designed and data were consecutively gathered with high granularity.
Limitations include that even though every effort was made to ensure that fluid input data were captured in a similar way at the different sites, we cannot exclude that smaller amounts of fluids were not registered or that subtle differences in collection of data may have contributed to the inter-site variability seen in the results. Also, we did not standardize registration of bowels movements in the fluid balance which may have resulted in an overestimation of fluid balance in some centres. Other limitations include the small sample size and that only two countries participated in the study. This may limit the validity of our findings in other countries.