In the absence of serious urgency for dialysis, the kidneys should be offered a chance to recover before the intensivist resorts to the use of dialysis. This strategy does not seem to change mortality and offers the opportunity for up to half of the patients to get away without any RRT whatsoever, without extending their ICU stay or hospital stay. Moreover it allows time for the patient to be pampered with the sort of refinements which only the ICU has to offer With careful haemodynamic resuscitation and biochemical correction a proportion of patients (those less sick at baseline) will probably never need dialysis, and be grateful for having fewer holes in their necks and groins.
These data have become available with the publication of the AKIKI trial (Gaudri et al, 2016) which seems like a ripe candidate for future written paper questions. The topic of RRT initiation had ultimately become a stem for a "critically evaluate" SAQ in Question 28 from the second paper of 2020. In a more viva-like setting, one might find people being presented with a somewhat borderline case scenario, and asked whether they would offer RRT (or wait).
Rationale for early commencement of renal replacement therapy
Generally, "early" or "prophylactic" RRT has been viewed as a strategy to minimise damage from the evil humours the accumulation of which results from renal insufficiency. The idea has been around since the 1950s. In brief point-form, the argument for removing these toxins early is fairly straightforward.
- Renal failure results in the accumulation of urea (which has certain physiological consequences) as well as non-volatile acids (which cause acidosis). Not to mention various unmentionable toxins which normally rely on the nephron for clearance.
- The damage done by this accumulation is dose-dependent, i.e. there is no "critical" value byond which the urea suddenly becomes toxic: it is toxic all the time, and the more of it there is the worse the patient's condition
- Ergo, the early removal of such toxins (without waiting for some sort of numerical threshold value) should benefit the patient by preventing complications of renal failure, such as fluid overload, metabolic encephalopathy, increased work of breathing due to acidosis, gastrointestinal haemorrhage, etc etc.
- For many patients presenting with renal failure, the need for future dialysis is apparent (i.e. you know you're going to dialyse them at some point) and it would make no sense to delay this therapy until specific criteria are met (eg. to wait for "symptomatic" uraemia).
- Delaying therapy is unlikely to have any immediate benefit as you have not treated anything.
- Medical therapies for managing complications of acute renal failure are not benign, for instance causing deafness (frusemide), bowel obstruction (cation exchange resins) and increased CO2 (bicarbonate).
Rationale for delayed commencement of renal replacement therapy
The reasons for delaying RRT are many-fold, but can be generally divided into three main groups: arguments from safety, arguments from pragmatism and arguments from physiology
Arguments from safety
- Renal replacement therapy is not benign: there are well-documented complications.
- A proportion of patients with severe acute kidney injury never go on to require RRT
- The safety of careful medical management is greater than the safety of careful RRT
- Specifically, the patient may not require central venous catheterisation, sparing them the risks of large vessel access, catheter-related blood stream infection, puncture site haematoma and air embolism.
Arguments from pragmatism
- Renal replacement therapy is not cheap.
- Most complications of acute renal failure are such that require relatively inexpensive and easily administered solutions, eg. calcium gluconate, soidum bicarbonate, frusemide, and so on.
- The cost of RRT is not only in equipment but also in manpower and maintenance.
- If spontaneous renal recovery is not achieved because RRT has caused some sort of "dialysis-induced renal injury", one may add the cost of more prolonged (potentially, life-long) dialysis.
Arguments from physiology
- Renal replacement therapy is not a "cure" for acute renal failure, it is merely a support strategy.
- RRT may even cause a dalay in renal recovery
- Some of the retained toxins of uraemia may have some opportunistic benefit, eg. urea may act as an osmotic diuretic (i.e. in the presence of a high urea the urine output may be greater).
Evidence in the literature
For exam purposes, this can be summarised as follows:
- Early RRT was popular in early 2000s. A good overview of Seabra et al (2008) were able to identify a nonsignificant 36% mortality risk reduction associated with early RRT, but lamented that early dialysis therapy looked best in the smaller studies, and that heterogeneity among the trials precluded definitive conclusions.
- Studies from that period were largely of poor quality, and were unable to find a mortality benefit. This early era is well summarised by Wierstra et al (2016) in their systematic review for Critical Care. The collaborators found nine "high quality" studies which were included in the final analysis, totalling 1042 patients. No survival benefit was seen.
- AKIKI trial - Gaudry et al (2016), with n= 620. Unlike the others, this was a randomised trial of early vs late initiation. Of the delayed group, only 51% received dialysis - the rest scraped through without it. There were also fewer catheter-associated infections in the delayed group. Sure, the ones which ended up being dialysed anyway looked biochemically worse when they received RRT, but they were already worse at baseline (with a higher SOFA score). Mortality between groups did not differ significantly (48.5% vs 49%), nor was there any increase in the length of ICU stay. On the basis of this trial, one may make a strong argument in favour of waiting for spontaneous renal recovery.
- ELAIN trial (Zarbock et al, 2016) - a smaller single-centre study which compares favourably in its methodology with AKIKI. Ther ELAIN investigators had an "early" group which was only separated from the "late" by a period of less than 24 hours on average. They also failed to treat the groups equally: the early group received RRT at KDIGO stage 2, but the delayed group had to wait untilt they got to Stage 3 before they got their dialysis. . The use of weird biomarkers (NGAL?) and unusual protocols (7 days of RRT in each group, regardless of renal performance) makes it difficult to use ELAIN as an argument for early dialysis. The ELAIN methodology is taken apart with appropriate viciousness in the JAMA editorial by Chertow and Winkelmayer, which brings up the question"if it's so crap then how come you published it?"
- STARRT-AKI (NEJM, 2020) was the last and arguably best volley in this exchange. As with everything CTG, it was massive (n = 2927) and showed no difference. Again, of the delayed group, a huge proportion (38%) never ended up getting dialysed. The early RRT group stayed in ICU for one day less (9 vs 10 days), and had a greater number of clinically trivial adverse events (hypophosphataemia, seriously). Interestingly, the early RRT group ended up having a much higher risk of dialysis dependence at day 90, which raises the question - does RRT, inflicted on the patient when they are at their most vulnerable, induce some sort of additional dialysis-induced kidney injury? Lasty, though informative, this trial did not help to determine what "delayed" really means for patients, and how to select the ones in whom delays are going to be harmful.
- AKIKI 2 (Gaudry et al, 2021) added some information to the latter. This was another large French study involving multiple centres and 278 patients. These oliguric patients resembled the delayed arm of STARRT-AKI, in the sense that they did not have any indications for immediate urgent CRRT. They were then observed for 72 hours of oliguria, or until their urea increased to over 40 mmol/L. At that stage they were randomised to either keep waiting or to receive RRT. It turned out that postponing RRT at this late stage did not conver any sort of benefit, and might have even been associated with harm (there was a statistically insignificant trend towards increased mortality).