Acute kidney injury after cardiac surgery (Thakar/Cleveland Clinic Score)

Clinical background

Dialysis-requiring acute kidney injury after cardiac surgery is uncommon, with an incidence of approximately 2 to 4 per cent in modern cohorts, but it carries a markedly increased mortality. The decision to operate on a patient with impaired renal function, or to choose between a surgical and a catheter-based strategy, is often weighed against this risk. Without a structured estimate, the assessment easily reduces to an intuitive weighting of age, renal function and comorbidity, in which the weights are arbitrary. The Thakar score, also known as the Cleveland Clinic score, was developed to combine the most important preoperative risk factors into a single weighted sum and thereby provide a reproducible basis for preoperative risk communication and resource planning.

Calculating the Thakar score

The score is a weighted sum of ten preoperative variables and ranges from 0 to 17 points:

$$\text{Score} = \text{female} + \text{heart failure} + \text{EF}<35$$

where type of surgery gives 0 points for CABG only, 1 for valve only, and 2 for CABG + valve or other cardiac surgery, and serum creatinine gives 0 points for < 1.2 mg/dL, 2 points for 1.2 to < 2.1 mg/dL, and 5 points for ≥ 2.1 mg/dL. The other binary variables give 1 point each, except a preoperative intra-aortic balloon pump and emergency surgery, which give 2 points each.

The score was derived at the Cleveland Clinic Foundation from 33,217 patients who underwent open cardiac surgery between 1993 and 2002 [1]. The model was developed on a randomly selected test set of 15,838 patients and internally validated on the remainder. The primary outcome was acute renal failure requiring dialysis, and the frequency of this outcome ranged from 0.5 per cent in the lowest risk category to 22.1 per cent in the highest [1]. The four risk categories, 0 to 2, 3 to 5, 6 to 8 and ≥ 9, were constructed arbitrarily on the basis of the observed rise in outcome frequency.

Interpretation in practice

The score translates into four risk bands, each with a concrete clinical meaning:

The score is a preoperative tool. It should not be used to guide postoperative management once kidney injury has occurred, and it does not replace clinical assessment of intraoperative factors such as bypass time and blood loss.

Validation and performance

In the internal validation in the derivation study, an AUC of 0.81 (95 per cent CI 0.78 to 0.83) was achieved in the test set and 0.82 (95 per cent CI 0.80 to 0.85) in the validation set, with no significant difference between the two [1].

Several external validations have confirmed the discriminatory ability of the score. In a Spanish multicentre cohort of 1,084 patients from 24 hospitals, in which 248 patients required renal replacement therapy, the Thakar score achieved an AUC of 0.82, outperforming the Mehta model (AUC 0.76) and comparable to the Simplified Renal Index (AUC 0.79) [2]. Calibration was, however, poor: the score tended to underestimate the actual need for renal replacement therapy [2].

In a Croatian cohort of 1,056 consecutive patients operated on between 2012 and 2014, with an incidence of dialysis-requiring kidney injury of 3.5 per cent, the Thakar score performed best of five models compared, with an AUC of 0.837 for dialysis-requiring kidney injury [3]. The score also discriminated reasonably for KDIGO-defined kidney injury of stage 2 or higher (AUC 0.811) [3].

In a Saudi single-cohort study of 329 patients, discrimination for dialysis-requiring kidney injury was lower, with an AUC of 0.717, and acceptable calibration (Hosmer–Lemeshow P = 0.342) [4]. The score slightly underestimated the incidence (predicted 1.7 per cent versus observed 2.1 per cent) [4].

An important limitation in the validation data emerged in a prospective Spanish multicentre study of elective cardiac surgery, in which emergency surgery and patients with an intra-aortic balloon pump were excluded [5]. In this population the discrimination of the Thakar score fell markedly to an AUC of 0.57 (95 per cent CI 0.50 to 0.64), indicating that the score loses much of its power when the most heavily weighted variables, emergency surgery and IABP, are absent [5].

Limitations

The score applies to open cardiac surgery on cardiopulmonary bypass. It is not validated for transcatheter procedures, off-pump surgery or minimally invasive surgery in the strict sense. Patients already on dialysis were excluded from the derivation cohort and the score should not be calculated for them.

The most heavily weighted variables — preoperative serum creatinine (up to 5 points), emergency surgery (2 points) and a preoperative IABP (2 points) — together account for a large part of the discriminatory ability of the score. In a population of elective patients only, without an IABP, the AUC falls dramatically, showing that the score is not a general tool for all cardiac surgery but depends on the high-risk patients being represented [5].

Serum creatinine as a risk marker is age- and sex-dependent, and the score uses fixed thresholds in mg/dL without correcting for eGFR. An older woman with low muscle mass may have a serum creatinine below 1.2 mg/dL despite impaired glomerular filtration and thus be underrated. Conversely, a younger man with a large muscle mass may be overrated.

Calibration is a recurring problem. Several external validations show that the score tends to underestimate the actual risk of dialysis-requiring kidney injury [2, 4], particularly in populations with a higher incidence than the derivation cohort. This means that the score may give false reassurance if used as the sole basis for forgoing enhanced postoperative monitoring.

The score does not include intraoperative factors such as bypass time, aortic cross-clamp time or blood transfusion, which several studies have shown to be independent risk factors for kidney injury [3, 4]. A patient with a low score but a long bypass time may carry a real risk that the score does not capture.

References

1. Thakar CV et al. A clinical score to predict acute renal failure after cardiac surgery. J Am Soc Nephrol 2005. PMID: 15563569
2. Vives M et al. External validation and comparison of three scores to predict renal replacement therapy after cardiac surgery: a multicenter cohort. Int J Artif Organs 2011. PMID: 21534243
3. Kristovic D et al. Cardiac surgery-associated acute kidney injury: risk factors analysis and comparison of prediction models. Interact Cardiovasc Thorac Surg 2015. PMID: 26091696
4. Alhulaibi AA et al. Validation of Various Prediction Scores for Cardiac Surgery-Associated Acute Kidney Injury. J Saudi Heart Assoc 2022. PMID: 36816793
5. Callejas R et al. Preoperative predictive model for acute kidney injury after elective cardiac surgery: a prospective multicenter cohort study. Minerva Anestesiol 2019. PMID: 29756690