As in the present study, a limited performance has been previously reported for pNGAL and CysC after cardiac surgery [
12]. In various settings, urine TIMP2 IGFBP7 is considered to be the most accurate biomarker [
19]. In cardiac surgery, a recent meta-analysis of 10 studies found that the pooled AUC
ROC of TIMP2 IGFBP7 for the detection of CS-AKI was remarkable (AUC
ROC = 0.83 [standard error 0.02]) [
9]. So why did not we find such good diagnostic accuracy for TIMP2 IGFBP7? Firstly, not all the available studies reported good performance for TIMP2 IGFBP7: some recent studies, not included in Su et al. meta-analysis, reported, as we did, a poor-to-moderate performance for TIMP2 IGFBP7 (AUC
ROC = 0.57–0.69) [
20‐
22]. Secondly, Day1 measurement of TIMP2 IGFBP7 has been included in this meta-analysis [
9]. For instance, in one study, an AUC
ROC of TIMP2 IGFBP7 at Day1 (0.71) has been retained whereas AUC
ROCs at the end of the surgery (AUC
ROC = 0.63) or at the 4th hour (AUC
ROC = 0.51) have not [
23]. However, beyond the fact that, at Day1, the superiority of TIMP2 IGFBP7 over pCr is highly uncertain, we believe that Day1 is a time point which is too late to provide nephroprotective interventions and is therefore not an
early diagnosis of CS-AKI. Thirdly, several studies defined CS-AKI only by pCr increase, i.e., omitted the urine output criterion [
24‐
29]. Yet, even possibly misleading, urine output is full part of the recommended definition of AKI [
4]. Last, in 3 studies included in Su et al.’s meta-analysis, TIMP2 IGFBP7 has been analyzed only for the detection of stage 2–3 CS-AKI [
9] and in most other studies, stage 2–3 CS-AKI featured prominently among cases of CS-AKI. This may have contributed to the good pooled performance of TIMP2 IGFBP7 for the detection of all stages CS-AKI. Indeed, the more severe the disease, the more sensitive and specific its biomarker. Hence, the diagnostic performance of TIMP2 IGFBP7 is probably better for the identification of moderate-to-severe CS-AKI (stage 2–3) than for mild CS-AKI (stage 1). It is noteworthy that TIMP2 IGFBP7 has been approved by the US Food and Drug Administration as an aid in the risk assessment for moderate or severe (stage 2–3) acute kidney injury within the next 12 h. Our study, in which most (92%) CS-AKI cases were mild (stage 1), contrary to several previous studies [
9], discourages to extend the indication for TIMP2 IGFBP7 to the detection of
mild CS-AKI developing within the 48 postoperative hours.
Our results challenges the long-held belief that pCr is too slow in its response to renal injury to provide a valuable prediction of adverse outcomes. However, our findings are in line with those of the scarce studies assessing the performance of early measurements of pCr. Indeed, minimal very short-term changes in pCr were associated with 30-day mortality after cardiac surgery [
30] or with a composite outcome of hospital mortality or need for RRT [
13]. Closer to the scope of the present study, i.e., early detection of CS-AKI, Ho et al. reported the utility of an early postoperative increase in pCr by ≥10% from baseline (odds ratio = 6.4 [95%CI:2.37–17.2]) [
31]. McIlroy et al. found that if early postoperative pCr measurements failed to decline, the occurrence of CS-AKI was likely (odds ratio = 2.6 [95%CI 1.7–4.1]) [
13]. These findings are reinforced by the present study, which, besides, provides the AUC
ROC for the CS-AKI outcome. We found the same threshold as that proposed by McIllroy et al. for change in pCr (0 μmol. L
− 1) [
13], in spite of the fact that we considered change in pCr from post-CPB to H6 (vs. from preoperative pCr to ≈ 3 h of ICU admission). Hence, minimal very short-term changes in pCr may reflect renal injury. Owing to perioperative volume expansion (especially during cardiac surgery with CPB), a decrease in pCr level is expected in the immediate aftermath of the procedure. Several hypotheses may be advanced if pCr fails to decline: reduced renal creatinine clearance, increased perioperative creatinine production, reduction in total body water (related to diuretic administration for instance) [
13,
31] or misleading measurements of pCr owing to analytical variability. However, the analytical precision of the enzymatic assay we used for pCr was good in our laboratory during the study period and beyond: maximum coefficient of variation of only 2.9% (even lower for elevated levels of pCr).
Study limitations
First, our findings observed in 65 patients, should be validated in a larger cohort. A larger study size is required for a more robust analysis of biomarkers combinations and a specific analysis of stage 2–3 CS-AKI detection. Additional studies will also be needed to assess whether a biomarker-guided implementation of nephroprotective measures at the 6th hour after the end of CPB (plus the turnaround time for pCr) positively impacts patients outcome. Of note, this has been found with TIMP2 IGBP7 measurements at the 4th hour [
36].
Second, owing to the exploratory nature of most analyses, we preferred preventing potentially useful findings not being overlooked [
32]. We therefore did not make adjustments for multiple tests but we cannot exclude that some significant differences we found were actually due to chance.
Third, only elderly patients (≥ 75 years-old) undergoing aortic valve replacement with CPB were included. Caution should therefore be exercised before any extrapolation to other populations. We chose to focus on this specific population because, in these often fragile patients, a less invasive alternative therapy for aortic valve stenosis is increasingly proposed when cardiac surgery with CPB is of questionable benefit/risk ratio: transcatheter aortic valve implantation [
34]. To optimize the safety of the open surgery option, reliable renal biomarkers are required to prompt nephroprotective measures [
4,
19].
Fourth, we cannot exclude that some false negative cases for TIMP2 IGFBP7 (or other biomarkers) were actually true negatives: some stage 1 CS-AKI cases could have been “functional”, i.e.
, with no kidney cell damage, explaining the lack of increase in novel biomarkers levels in patients with mild increase in pCr and/or oliguria [
9]. However, “
even mild reversible AKI has important clinical consequences” as stated in KDIGO guidelines and supported by several studies in the setting of cardiac surgery [
1,
2,
12,
13,
30,
33,
35], and should therefore be detected as early as possible [
4,
33].
Fifth, we did not strictly use all the criteria of the KDIGO definition. Indeed, we focused on the renal impact of the index cardiac surgery procedure (and very early potential complications) rather than later complications. As other authors [
1,
3,
36], we therefore arbitrarily reduced the observation period of the CS-AKI definition. We used a 48 h-time window for all criteria although one of the three criteria of the KDIGO definition (relative increase in pCr) has been proposed with a 7-day time window [
4]. Enlarging the time window, i.e., using a more sensitive definition, could have yielded higher incidence of CS-AKI and, possibly, different performance of the tested biomarkers.
Last, pCr was both biomarker under-test and full part of the definition (within 48 postoperative hours, along with oliguria) of the outcome measure (CS-AKI). Although only early (H6 or earlier) measurements of pCr were analyzed for the early detection of CS-AKI, pCr may therefore be seen as both judge and judged. Again, we used the recommended definition of AKI [
4]. Confirming our findings by using a definition of CS-AKI fully unrelated to pCr would be desirable, but such robust standard does not exist to date.