AKIN classification for acute kidney injury

Clinical background

The AKIN classification for acute kidney injury was developed to create a uniform definition and staging of acute kidney injury (AKI), a condition that long lacked accepted diagnostic criteria and in which different studies used different thresholds for creatinine and urine output. Its predecessor, the RIFLE classification (2004), was a first step, but data showing that even small rises in creatinine, of the order of 0.3 mg/dL (26 µmol/L), were associated with increased mortality prompted a revision [1]. The AKIN classification therefore introduced an absolute creatinine criterion for stage 1, limited the diagnostic window to 48 hours, and classified all patients starting renal replacement therapy as stage 3 irrespective of other values [1].

The tool serves two decisions: whether the patient has AKI at all, and how to stage the severity of the injury for prognosis and choice of monitoring level. The decision is difficult without a structured criterion, because creatinine is a late and non-linear marker of glomerular filtration, and urine output is influenced by volume status, diuretics and obstruction.

Applying the AKIN classification

AKIN rests on two variables: the change in serum creatinine relative to the baseline creatinine, and urine output. The diagnosis requires the change to occur within 48 hours. Staging follows the highest criterion met:

$$\text{Stage 1: } S_{Kr} \times 1{.}5\text{--}1{.}99 \text{ or } \Delta S_{Kr} \geq 0{.}3 \text{ mg/dL, or UO} < 0{.}5 \text{ mL/kg/h for } > 6 \text{ h}$$

$$\text{Stage 2: } S_{Kr} \times 2\text{--}2{.}99 \text{, or UO} < 0{.}5 \text{ mL/kg/h for } > 12 \text{ h}$$

$$\text{Stage 3: } S_{Kr} \times \geq 3 \text{, or } S_{Kr} \geq 4{.}0 \text{ mg/dL with an acute rise } \geq 0{.}5 \text{, or UO} < 0{.}3 \text{ mL/kg/h for } \geq 24 \text{ h / anuria } \geq 12 \text{ h, or ongoing RRT}$$

where $S_{Kr}$ is the current serum creatinine, $\Delta S_{Kr}$ is the absolute rise from baseline, UO is urine output and RRT is renal replacement therapy. The final stage is the highest criterion met, whether it derives from the creatinine or the urine output criterion.

The derivation is a consensus report from the Acute Kidney Injury Network, produced at a conference in Amsterdam in September 2005 with representatives of intensive care and nephrology societies [1]. It is not derived from a specific patient cohort but is a modification of the RIFLE criteria supported by epidemiological data showing that small rises in creatinine were associated with adverse outcomes. The consensus group explicitly noted that the criteria needed validation in future studies [1].

Interpretation in practice

An important principle in AKIN is that renal replacement therapy is not a stage in its own right but an outcome that places the patient in stage 3 [1]. This means that a patient who starts dialysis has stage 3 by definition, however modest the creatinine rise.

Validation and performance

Several external validation studies have compared AKIN with RIFLE and with the later KDIGO classification (2012).

In a German single-centre study in a general intensive care unit with 321 patients, AKIN and RIFLE classified 86.8 per cent of patients into concordant severity grades [2]. AKIN identified more patients with AKI than RIFLE (38.6 per cent versus 32.4 per cent), chiefly through the absolute creatinine criterion for stage 1. The AUC for predicting 28-day mortality was 0.73 for AKIN compared with 0.71 for RIFLE, while the unclassified maximum serum creatinine performed at least as well, with an AUC of 0.76 [2]. For predicting the need for renal replacement therapy the AUC was 0.83 for both AKIN and RIFLE [2].

In a large Japanese retrospective study of 49,518 hospital admissions, Fujii et al. found that AKIN identified considerably fewer patients than either RIFLE or KDIGO: only 4.8 per cent with AKIN compared with 11.0 per cent with RIFLE and 11.6 per cent with KDIGO [3]. Discrimination for in-hospital mortality was significantly poorer for AKIN (AUC 0.69) than for RIFLE (AUC 0.77, P<0.001) and KDIGO (AUC 0.78, P<0.001) [3]. The lower incidence is explained by the 48-hour window of AKIN, which excludes patients with a slower creatinine rise that RIFLE and KDIGO capture within their 7-day window.

In a prospective multicentre study of 3,107 intensive care patients at 30 ICUs in Beijing, Luo et al. found that AKIN identified AKI in 38.4 per cent of patients, compared with 46.9 per cent for RIFLE and 51.0 per cent for KDIGO [4]. The AUC for in-hospital mortality was 0.746 for AKIN, 0.738 for RIFLE and 0.757 for KDIGO, with no significant difference between AKIN and KDIGO (P=0.38) [4]. The 391 patients whom KDIGO identified but AKIN missed nonetheless had significantly higher mortality than patients without AKI, indicating that the 48-hour restriction of AKIN produces false negative classifications of clinical relevance [4].

In summary, AKIN performs comparably to KDIGO in intensive care patients, but less well in a general hospital population, in which its lower sensitivity becomes more apparent.

Limitations

The 48-hour window is the single most important limitation. AKIN requires the change to occur within 48 hours, which excludes patients with a progressive creatinine rise over a longer period. This is the primary reason why AKIN identifies fewer patients than KDIGO, which allows a 1.5-fold rise within 7 days [3, 4]. Patients missed by AKIN but captured by KDIGO have been shown to have higher mortality [4].

The baseline creatinine must be known. Where it is not, AKIN recommends assuming normal renal function, which leads to two systematic errors: patients with unrecognised chronic kidney disease may be misclassified as having AKI (type I error), and patients with a low baseline creatinine may miss the diagnosis of AKI (type II error) [2]. In an intensive care population, in which the baseline is often unknown, this is a substantial problem.

The urine output criterion is sensitive to volume status, diuretics and obstruction. A minority of the consensus group questioned at the time of derivation whether a urine output below 0.5 mL/kg/h over six hours is specific enough to diagnose AKI reliably [1]. In addition, hourly urine measurement is not routine outside intensive care units, which limits its usefulness on general wards.

AKIN does not apply to patients on chronic dialysis or to kidney transplant recipients, and the criteria have not been validated in children. The consensus group also noted that the criterion of an absolute rise of 0.3 mg/dL needs separate validation for patients with pre-existing chronic kidney disease, in whom a similar absolute rise from an already raised baseline may carry a different prognostic significance [1].

KDIGO has largely replaced AKIN in clinical practice since 2012. KDIGO combines the strengths of both RIFLE and AKIN, allows both a 48-hour and a 7-day window, and has shown better or equivalent performance in all comparative studies [3, 4]. AKIN nonetheless retains value as a conceptual framework and as a reference in the older literature.

References

1. Mehta RL, Kellum JA, Shah SV, et al. Acute Kidney Injury Network: report of an initiative to improve outcomes in acute kidney injury. Crit Care 2007;11(2):R31. PMID: 17331245
2. Huber W, Schneider J, Lahmer T, et al. Validation of RIFLE, AKIN, and a modified AKIN definition ("backward classification") of acute kidney injury in a general ICU: Analysis of a 1-year period. Medicine (Baltimore) 2018;97(38):e12465. PMID: 30235738
3. Fujii T, Uchino S, Takinami M, et al. Validation of the Kidney Disease Improving Global Outcomes criteria for AKI and comparison of three criteria in hospitalized patients. Clin J Am Soc Nephrol 2014;9(5):848–854. PMID: 24578334
4. Luo X, Jiang L, Du B, et al. A comparison of different diagnostic criteria of acute kidney injury in critically ill patients. Crit Care 2014;18(4):R144. PMID: 25005361