CRRTnet: a prospective, multi-national, observational study of continuous renal replacement therapy practices

Acute kidney injury (AKI) occurs commonly in critically ill patients, complicating more than half of intensive care unit (ICU) admissions [1, 2]. About 5–6% of ICU patients will receive renal replacement therapy (RRT), and hospital mortality rates in this population approach or exceed 50% [3‐6]. Despite the major clinical significance of this syndrome, there remains no proven therapy to reverse or attenuate established AKI, and care is primarily supportive, including RRT in severe cases.

For ICU patients, particularly those with hemodynamic instability, continuous renal replacement therapy (CRRT) has emerged as the dialysis modality of choice in high-income nations. Since the original descriptions of continuous arteriovenous hemofiltration over 30 years ago, [7] CRRT has undergone significant evolution with several improvements allowing more widespread and routine use. Advancements include peristaltic pump driven venovenous circulation, sophisticated hardware and software engineering for precise application of the prescribed treatment and enhanced safety and alarms, improved biocompatibility and other membrane characteristics, regional anticoagulation protocols and enhancements in user interface. Indeed, for hemodynamically unstable patients the 2012 Kidney Disease: Improving Global Outcomes (KDIGO) AKI guidelines suggest using CRRT over intermittent dialysis options [8]. Compared to standard intermittent hemodialysis (IHD), CRRT offers advantages of improved hemodynamic stability, better overall solute clearance, and better fluid balance [8‐10]. Observational studies also suggest higher rates of renal recovery among patients initially treated with CRRT compared to conventional IHD, although this benefit has not been observed in randomized trials [11].

While the clinical role of CRRT has become well-established, there exists tremendous variability in how CRRT is prescribed and delivered [12‐14]. The KDIGO AKI guidelines provide some recommendations, stating that the target delivered CRRT effluent dose should be 20–25 mL/kg/h, and also suggesting that regional citrate should be the first-line form of anticoagulation [8]. However, specific recommendations are lacking for a number of other important areas, such as CRRT modality, details of anticoagulation protocols, electrolyte supplementation practices, or approach to fluid balance. This reflects the lack of strong evidence upon which to base guidelines, as most studies examining CRRT practices have been relatively small and/or single-center. In addition, despite the high-risk nature of the patient population, very few studies have examined quality metrics and/or patient safety related to CRRT practices [15].

Here we describe the protocol for a prospective observational study focused on CRRT practices in critically ill adult patients across a network of medical centers in North America (CRRTnet). Our goal is to develop a large repository of data regarding CRRT practices and outcomes, allowing characterization of similarities and differences in standard CRRT practices across medical centers. The natural variability in practices between (and even within) centers will provide the ability to compare outcomes between different practice patterns and generate important preliminary data to inform future interventional trials. In addition, a long-term goal for CRRTnet is to establish benchmarks for best CRRT practice for centers both within and outside the study.

Severe AKI remains a common complication of critically ill patients, and recent studies suggest that the incidence of dialysis-requiring AKI is increasing [18]. CRRT has emerged as the recommended dialysis modality for hemodynamically unstable patients in the ICU, and a recent study demonstrates that CRRT is the most commonly used dialysis modality in critically ill patients with AKI [19]. CRRT provides the advantages of improved solute and fluid balance control compared to patients on intermittent therapies. Yet in reality “CRRT” is an umbrella term that encompasses multiple different modalities and approaches. A high degree of variation exists in how CRRT is prescribed and delivered between (and within) medical centers, and there is currently a paucity of high-quality evidence to guide best practices. CRRTnet was developed with the goal of establishing the largest registry of CRRT patients to date, allowing for comparison of outcomes between a variety of different CRRT practices.

Ideally, each aspect of therapy would be assessed in randomized clinical trials. However, given the complexities of critically ill patients and potential difficulties with recruitment in this population, such an approach may not be feasible to assess every aspect of CRRT care. As such, observational studies remain an important tool, particularly when significant variations in practice naturally exist. As an example, CRRTnet was modeled after the Prospective Pediatric CRRT registry (ppCRRT) [20]. Over a 4 year period and including 13 centers in the United States, the ppCRRT enrolled 370 critically ill children who underwent CRRT and characterized outcomes in this previously under-described population [21]. Subsequent studies from ppCRRT explored patient outcomes related to vascular access, [22] anticoagulation, [23] and nutritional practices [24]. These studies provided important insight into CRRT practices and outcomes in children, spurred additional research, and continue to guide best practices in this population by informing the KDIGO AKI guidelines. Although there is significantly more literature regarding CRRT practices in adults, most of this literature is comprised of relatively small studies, primarily from single-centers. As such, there remains a significant need for large, multi-center studies such as CRRTnet.

Beyond descriptive analyses of patient characteristics and overall outcomes, we are planning on several comparative analyses. One particular area of interest is in the management of fluid overload. Our data will allow determination of fluid overload status at time of CRRT initiation, as well as daily fluid balances thereafter. While the dangers of established fluid overload have been increasing recognized in recent years, [10, 25, 26] there is a paucity of data examining different approaches to fluid management utilizing renal replacement therapy. We will examine rates of fluid removal (e.g. in ml/kg/h) and fluid overload correction to determine potential impact on patient outcomes such as renal function recovery and mortality. We hypothesize that, after adjusting for severity of illness, higher fluid removal rates will be associated with lower likelihood of renal recovery.

Another area of particular interest is anticoagulation to maintain CRRT filter patency. The KDIGO guidelines suggest regional citrate anticoagulation as the first-line method in patients without a contraindication. We anticipate that citrate anticoagulation will be associated with longer circuit life compared to heparin or no anticoagulation, a finding that has been observed in a number of clinical trials [27‐30]. However, there is no consensus approach to citrate anticoagulation; while several different protocols have been published, there have been no comparative trials. Because of existing differences in citrate anticoagulation protocols among CRRTnet centers, we will be able to compare outcomes between different protocols.

One novel aspect of CRRTnet is the capture and analysis of machine data. These data record all machine measurements to the second, and include pressure measurements (e.g. access and return line pressures, filter pressures, transmembrane pressures), alarms and response to alarms, any changes in prescription (including blood, dialysate and replacement flow rates), interruptions, and actual delivered therapy. To our knowledge, this is a relatively untapped source of information, and one goal of CRRTnet will be to develop algorithms for processing the information downloaded from the CRRT machine memory cards. One of the challenges of CRRT is the complexity of therapy and the increased associated workload compared to intermittent therapies [31]. By utilizing the detailed treatment data, we will be able to characterize alarm frequency and provide more precise estimates on interface time required to maintain this therapy. This information may be helpful to administrators in determining nursing models and CRRT educational programs. In addition, nursing workload can be considered an important secondary outcome that may vary between and within centers depending on variations in CRRT protocol. For example, we hypothesize that regional citrate anticoagulation will be associated with less workload compared to non-citrate anticoagulation. It will also be interesting to compare associated workload between different citrate protocols, especially if circuit filter life is similar.

An important strength of our study is that we are examining everyday practices. While findings from controlled clinical trials can sometimes be difficult to duplicate in the “real world” setting, we will provide a relatively unfiltered look. This may be particularly important when examining adverse events and complications of therapy. Conversely, it is important to recognize that CRRT is performed at a relatively high frequency at each of the study sites, which facilitates development of local expertise and maintenance of competency. Whether or not similar practices will result in similar outcomes at smaller centers with lower volumes is uncertain. Nonetheless, we believe that CRRTnet will provide important outcomes benchmarks for centers performing CRRT.

Another important strength of CRRTnet is the ability, through waiver of consent, to include all eligible participants. In terms of the practicality of obtaining informed consent for an observational study in a population similar to our proposal, much can be learned from the unfortunate experience of the PICARD study investigators where patients and substitute decision makers were approached for informed consent [32]. Enrollment was poor, with only 52% of eligible patients included, greatly hindering the generalizability of the results. Patient refusal was infrequent at all sites, suggesting that patients themselves did not have concerns or objections to having their data collected when they were able to be approached directly. However, refusal by family occurred in almost one fifth of instances, and the absence of family or appropriate proxy was listed as a cause for refusing enrollment in as high as 40% at one site. Finally, many patients could not be enrolled because of death or discharge before being seen by study personnel. Thus, countless resources were spent on conducting an observational trial that missed almost half the eligible subjects, whose data very likely would have influenced the results of the study were they included. The CRRTnet database will be highly generalizable based on the inclusion of most eligible patients.

Some limitations of our study design are worth noting. First, we will truncate CRRT data at 14 days, which will limit applicability in patients on prolonged courses of CRRT. This decision was made primarily as a feasibility measure to limit the workload associated with detailed clinical data collection; importantly, preliminary data from participating centers showed that median duration of CRRT was in the range of 6–8 days, so we will still capture the full CRRT course for the majority of patients enrolled. Second, we will not be enrolling patients with AKI who only receive intermittent dialysis therapies and we recognize that there can be significant variation in which patients are chosen to undergo CRRT versus intermittent therapies. However, it is important to note that our goal is not to establish the role of CRRT in AKI management, but rather to focus on practice variation within CRRT prescription and delivery. So while we will not be able to advocate for which patients should be managed with CRRT, we hope to inform clinicians of best practices once a decision to initiate CRRT is made.

Over the past 30 years, CRRT has developed into an important tool for the management of critically ill patients with renal failure. Despite overall improvements, there remains marked variation in CRRT practices which is driven in part by lack of evidence-based guidance. The primary goal of CRRTnet is to develop a large data repository of CRRT patients with detailed information regarding specific CRRT practices which can then be compared in relation to outcomes. We anticipate that CRRTnet will provide the nephrology and critical care communities with an important resource for planning future clinical trials as well as for benchmarking clinical outcomes.

Recruitment is currently active at the 4 medical centers mentioned above. A fifth center is undergoing evaluation to join CRRTnet. The initial planned recruitment phase is 5 years with a target enrollment of 2000 patients.