Outcomes of sustained low efficiency dialysis versus continuous renal replacement therapy in critically ill adults with acute kidney injury: a cohort study

Patients were identified from the St. Michael’s Hospital Acute Kidney Injury Registry, which comprises all adults who commenced RRT for AKI in the hospital’s four intensive care units (medical-surgical, cardiovascular, trauma-neurosurgical or coronary care), between April 2007 and September 2012. AKI was defined by the Acute Kidney Injury Network (AKIN) criteria [14]: minimum creatinine increase of 27 μmol/L or 50 % rise from baseline (if baseline ≥ 354, then a rise of 44 μmol/L qualified for a diagnosis of AKI). Patients with pre-existing end-stage renal disease who were erroneously entered in the Registry were excluded. We also excluded patients who initiated RRT using intermittent hemodialysis (IHD) because at our institution, IHD is reserved for patients who are hemodynamically stable whereas SLED or CRRT are deployed in patients with perceived hemodynamic instability. We restricted this analysis to patients for whom we could ascertain vital status at 30 days following the initiation of RRT.

Our study was approved by the Research Ethics Board of St. Michael’s Hospital. The need for patient-level consent was waived by the Research Ethics Board.

Initial RRT modality, SLED versus CRRT, was the exposure of interest. Choice of initial RRT modality was made at the discretion of the consulting nephrologist in collaboration with the attending critical care physician. In the absence of compelling clinical or personnel concerns, hospital policy discouraged switches between CRRT and SLED.

CRRT was administered by ICU nurses as continuous venovenous hemodiafiltration or continuous venovenous hemofiltration using Prisma and Prismaflex (Gambro, Richmond Hill, ON) devices. AN-69-based filters were utilized, with blood flows ranging from 100 – 200 mL/hr and target effluent rates of 20 – 35 mL/kg/hr.

SLED was introduced as an alternative to CRRT in 2007. Both modalities were targeted to hemodynamically unstable patients. SLED was delivered by dialysis nurses using Phoenix^TM dialysis machines (Gambro, Richmond Hill, ON) and CA210 or Xenium 210 dialyzers (Baxter, Deerfield, IL). SLED sessions were targeted to 8 h in duration (minimum 6 h) with blood and dialysate flows of 200 and 350 mL/min, respectively. The minimum frequency of SLED treatments was three times per week but could be increased at the nephrology team’s discretion.

The nephrology consultation service, comprising trainees who were supervised by an attending nephrologist, prescribed all CRRT and SLED treatments.

We obtained the following data from the St. Michael’s Hospital Acute Kidney Injury Registry: patient age and gender; reason for ICU admission; pre-morbid kidney function (derived from the abbreviated Modification of Diet in Renal Disease formula [15] using the last available pre-hospitalization serum creatinine); Charlson comorbidity score [16]; and the following parameters at the time of RRT initiation: basic biochemistry and hematologic values; urine output, receipt of mechanical ventilation, requirement for vasopressors and the Sequential Organ Failure Assessment (SOFA) score [17] (as modified for use in the Registry [18], see Additional file 1) at the time of RRT initiation. The SOFA score was also recorded at 48 h following RRT initiation.

The primary outcome was all-cause mortality at 30 days following RRT initiation. Secondary outcomes included RRT dependence at 30 days, cumulative fluid removal seven days following initiation of RRT, and early clinical deterioration, defined as death or increase in SOFA score within 48 h of RRT initiation)

We expressed continuous variables as means [standard deviations (SD)] or medians [interquartile range (IQR)], as appropriate, and categorical variables as numbers (percentages). The characteristics of patients who initiated SLED vs CRRT were compared using the t-test, Mann-Whitney U test or Fisher’s exact test. Logistic regression was used to evaluate the relationship between RRT modality (SLED versus CRRT) and the outcomes of interest. For the primary outcome of mortality at 30 days, we adjusted for age, sex, ICU type, Charlson score, mechanical ventilation and vasopressor status, serum creatinine at the time of RRT initiation, SOFA score and urine output at the time of RRT initiation. These covariates were chosen based on clinical relevance. For the secondary outcomes for which there were fewer events, we adjusted our models for baseline serum creatinine, SOFA score, urine output at the time of RRT initiation and Charlson score.

We examined the primary outcome in a subgroup analysis wherein we stratified patients by initial SOFA score (less than vs greater than or equal to the median value of 16) to determine whether severity of illness modified the relationship between modality and 30-day mortality. Since switches in RRT modality (eg, SLED to CRRT) may misclassify the exposure and dilute the association between initial RRT modality and clinical outcomes, we conducted two sensitivity analyses. In the first, we designated RRT modality as the one used for the majority of treatment sessions received from the initiation of RRT through Day 30. In the second, we restricted the cohort to patients who remained on a consistent modality for the first three RRT sessions.

We performed all analyses using SAS software version 9.1.3 (SAS Institute Inc., Cary, NC).

Baseline characteristics, stratified by initial receipt of SLED or CRRT, are presented in Table 1. Patients who initiated CRRT had higher Charlson (2.61 ± 2.53 versus 1.79 ± 1.77 for SLED recipients, p = 0.03) and SOFA(16.4 ± 3.08 versus 15.4 ± 3.65 for SLED recipients, p = 0.03) scores at the time of RRT initiation and were more likely to be mechanically ventilated (96.5 versus 85.6 % of SLED recipients, p = 0.03).

We examined 1107 treatment sessions (698 CRRT and 509 SLED sessions) to characterize the RRT delivered to members of the cohort. The median duration of SLED sessions was 7.11 h (IQR 6.00–7.92). On average, 91.9 % (SD 1.7 %) of the prescribed SLED treatment duration was delivered. The mean proportion of prescribed CRRT time that was actually delivered was 84.8 % (SD 22.0 %), which is compatible with 20.3 h in a 24-h period. SLED was delivered without use of anticoagulation in 86 % of treatment sessions, compared to 6 % of CRRT sessions. Further information on the delivered therapies is shown in Table 2.

All-cause mortality at 30 days was 54 % and 61 % among SLED- and CRRT-treated patients, respectively. After adjustment for possible confounders, SLED was not associated with 30-day mortality (adjusted OR 1.07, 95 % CI 0.56–2.03), as compared with CRRT (Table 3, Fig. 2). Additionally, there were no differences in secondary outcomes between the groups, including the adjusted risk of RRT dependence at 30 days and clinical deterioration within 48 h of RRT initiation (Table 4). Cumulative fluid removal at seven days was not significantly different between the groups (5846.5 ± 8855.6 mL versus 8180.2 ± 11322.0 mL, respectively).

There was no association between RRT modality and mortality within strata of acute illness severity (SOFA ≥ vs. < 16, Table 4).

We repeated the primary analysis with re-categorization of patients according to the RRT modality received for the majority of treatment sessions during the first 30 days. SLED was not significantly associated with 30-day mortality [adjusted OR 1.70, 95 % CI 0.76–3.28]. Similarly, after restricting our cohort to patients who received a consistent modality during the first three treatment sessions (n = 183), we did not identify a relationship between SLED and 30-day mortality [adjusted OR 1.36, 95 % CI 0.61–3.82] (Table 4).

It has been hypothesized that after AKI, autoregulation of blood flow is impaired, with susceptibility to further kidney damage as a result of dialysis-related hemodynamic instability [22, 23]. CRRT has the purported benefit of improved hemodynamic tolerability compared to IHD [1, 2, 24]. Long-term kidney outcomes may also be better among patients who received CRRT [24‐26]. Although we demonstrated that surviving SLED recipients are not more likely to remain RRT-dependent as compared to CRRT recipients, our sample size is small and the relative impact of SLED on long-term kidney function is an important question for future study.

Our findings also demonstrated comparable volume control among SLED and CRRT recipients in the week after RRT initiation and support the findings of previous smaller observational studies [7, 27].

Since the need for anticoagulation often presents a series of practical challenges in critically ill patients who require RRT, our finding that the vast majority of SLED sessions could be delivered with no anticoagulation is compelling. Since clinical outcomes of SLED seem comparable to CRRT, the ability to deliver RRT without the bleeding and metabolic complications of current anticoagulation options commonly used in CRRT (e.g., heparin or regional citrate anticoagulation) may represent a major benefit of SLED. Delivery of SLED predominantly without anticoagulation has not been reported in other studies assessing this modality [13, 19, 20, 28].

Our study’s strengths include use of a robust database that comprises clinically relevant measures of comorbidity, baseline renal function, biochemical and hemodynamic indices, and acute illness severity. This enabled us to adjust for critical confounders of the relationship between RRT modality and outcomes and at least partially overcome bias related to factors that may influence a clinician to initiate RRT with either SLED or CRRT. Our study included assessment of early clinical deterioration (within the first 48 h after RRT initiation), which, to date has not been examined in the literature and showed that SLED recipients did not experience a more rapid clinical decline during the critical period after RRT initiation. Our study was also conducted in a centre where both SLED and CRRT were readily available treatment options. This allowed us to compare contemporaneous patients, rather than historical controls, as occurs in situations when hospitals make wholesale switches from one modality to the other. Finally, given that our study is, to our knowledge, the largest non-randomized comparison of SLED and CRRT and comprised critically ill patients with AKI from a variety of causes, we expect that our results will have broad generalizability and complement information gleaned from previous smaller studies. As such, we believe that this observational study provides vital “real world” information that encompasses a broader spectrum of patients that might be enrolled in clinical trials.

Our study also has evident limitations. As a non-randomized study, our findings are susceptible to unmeasured confounding. Given the greater pre-existing experience and comfort level with CRRT, it is possible that patients with more severe illness were treated with CRRT preferentially over SLED. This is suggested by the slightly higher SOFA and Charlson scores and the higher frequency of mechanical ventilation in the CRRT group. These statistically significant differences may bias clinical outcomes in favour of SLED. In order to mitigate the impact of confounding, we included multiple measures of pre-morbid health and acute illness severity in our multivariable model. After adjustment for these imbalances, the receipt of SLED was not associated with inferior outcomes. The primary analysis examined patients based on the initial RRT modality received; however, we recognized that patients could have been treated with both modalities, which would bias our results towards the null hypothesis (or no effect). It was therefore reassuring that our sensitivity analyses were consistent with our primary analysis. Our study was restricted to 30-day follow-up and as a result, we were unable to comment on mortality or RRT dependence associated with either modality over the long-term. It has been argued that the nephroprotective benefits of CRRT only become evident 90 days or more after initiation of RRT, once the acute illness has abated [25, 26]; however such long-term data on SLED-treated patients remains limited. Finally, although the current study is among the largest to compare SLED and CRRT with respect to clinical outcomes, our sample size is still too small to definitively evaluate whether SLED is non-inferior to CRRT.