Serum Cystatin C as an Early Predictive Biomarker for Acute Kidney Injury (AKI) in Neonates: A PRISMA-DTA Systematic Review

1. Introduction

Acute kidney injury is an important complication among the neonatal population inside the neonatal intensive care unit (NICU). It is a significant cause of morbidity and mortality, therefore early detection is essential for timely intervention [1,2,3]. Nowadays, the assessment of AKI in neonates is challenging since the diagnosis relies primarily on traditional biomarkers, which include serum creatinine (sCr) levels, and urine output (oliguria/anuria) [4,5]. Currently, neonatal AKI is classified using standardized systems such as the neonatal-modified KDIGO (Kidney Disease: Improving Global Outcomes), AKIN (Acute Kidney Injury Network), and pRIFLE (pediatric Risk, Injury, Failure, Loss, End-Stage Renal Disease). [6,7] These systems define and stage AKI severity based on the magnitude of sCr increase from baseline and/or the duration of diminished urine output [6,7,8,9].

However, the traditional biomarkers are limited by multiple disadvantages such as:

-

maternal influence: in the first 24 to 72 hours of life, neonatal sCr represents maternal renal function and not the infants’, as creatinine freely crosses the placenta;

-

gestational age dependence: sCr levels decline and stabilize at varying rates over the first few weeks of life, meaning its baseline value is highly affected by gestational age (GA);

-

delayed detection: sCr is a late AKI marker as it measures function rather than actual tissue damage; sCr often does not rise until 48 to 72 hours after the renal injury, typically its levels rise only after almost half of the nephrons have already been compromised;

-

non-renal factors: sCr production and concentration depend on variables such as muscle mass, hydration status, age, and gender. [10,11,12,13]

Recent research has focused on serum cystatin C (sCysC) as a potential early biomarker for AKI in neonates. Serum Cystatin C is a 13.3 kDa cationic protein produced at a constant rate by all nucleated cells. Due to its small size, sCysC is freely filtered through the glomerular membrane and is subsequently completely reabsorbed and catabolized by the proximal renal tubules, without undergoing tubular secretion [13].

Serum Cystatin C has the following advantages over other biomarkers: it is minimally affected by muscle mass and maternal renal function, it is not significantly influenced by gender, gestational age, muscle mass and hydration status, and reacts rapidly to alterations in the glomerular filtration rate (GFR). [4,13,14] In order to achieve an early diagnosis, the use of sCysC instead of sCr in the neonatal population presents fundamental physiological advantages.

Despite no current consensus, multiple reviews and primary clinical studies suggest that sCysC may be an effective early biomarker for AKI in neonates, owing to its higher sensitivity than traditional biomarkers. [6,15,16,17,18,19] There is a growing interest in sCysC as a biomarker for neonatal AKI; however, essential aspects remain insufficiently clarified. The most important ones are when, how, and in which neonatal populations sCysC should be used. There are currently no standard threshold values for sCysC to accurately diagnose neonatal AKI. Another problem is the fact that most of the studies were conducted on a mixed population, both full-term and preterm neonates, who have different degrees of renal maturation.[16] Moreover, AKI diagnosis is usually established using modified neonatal KDIGO, AKIN, or pRIFLE criteria, but the reference standards (including measurement time points) vary considerably across studies.[20]

Previous reviews pooled data from studies with heterogeneous time points of sCysC measurements [16.17]. Among those reviews there was no distinction between the biomarker’s capacity to predict AKI before clinical manifestation and its ability to confirm AKI concurrently diagnosed with creatinine-based criteria. As the clinical implications of these two biomarkers are fundamentally different, this represents a gap in the literature.

Consequently, an important question remains unanswered: does sCysC perform better as an early predictor of neonatal AKI or as a concurrent diagnostic marker?

This review investigated the effectiveness and the role of sCysC in early AKI detection in newborns. We attained a structured narrative synthesis by systematically analysing the existing studies and separating early predictive performance from concurrent diagnostic values of this biomarker.

2. Patients and Methods

A. Protocol and Registration: this systematic review was conducted and reported in accordance with Preferred Reporting Items for Systematic Reviews and Meta-Analyses for Diagnostic Test Accuracy (PRISMA-DTA) guidelines. The completed PRISMA-DTA checklist is provided in Supplementary Material 1. Additionally, the narrative synthesis adhered to the Synthesis Without Meta-analysis (SWiM) reporting guidelines (Supplementary Material 2.) The protocol was prospectively registered in PROSPERO International Prospective Register of Systematic Reviews (Registration ID: (CRD420261302574).

B. Eligibility criteria

Studies were eligible for this systematic review if they met the following criteria:

a) Population: neonates from birth to 28 days of postnatal age. Both term (≥37 weeks) and preterm (<37 weeks) neonates across all clinical settings were considered (e.g., NICU, cardiac surgery). A minimum sample size of 10 neonates was required. This threshold was settled to minimize the risk of small studies effects. Exclusion criteria involved: mixed pediatric populations without extractable neonatal data; primary congenital kidney anomalies; studies focused exclusively on chronic kidney disease.

b) Index test: serum cystatin C (sCysC) measured by any validated method (e.g., PENIA, PETIA, ELISA). Given the potential validation in calibration between different assays, the specific measurement method was extracted from each study and treated as a potential source of methodological heterogeneity. Studies assessing only urinary cystatin C or reporting only cystatin C–based estimated glomerular filtration rate (eGFR) without extractable data for diagnostic accuracy were also excluded.

c) Reference standard: the reference standard used for AKI diagnosis was defined by validated sCr and/or urine output criteria, including the neonatal-modified KDIGO, original KDIGO, AKIN, RIFLE, or pRIFLE criteria. Baseline sCr was required to be defined using postnatal measurements obtained after the initial period of maternal creatinine influence (typically after 24-48 hours of life).

c) Target condition: neonatal AKI

d) Study design: Prospective and retrospective cohort studies, cross-sectional diagnostic accuracy studies reporting sensitivity, specificity, AUC-ROC (area under the receiver operating characteristic curve), cut off value.

e) Outcomes: diagnostic accuracy measured through various metrics including sensitivity, specificity, AUC-ROC and optimal threshold values

C. Information sources and search strategy:

a) Information Sources Databases:A comprehensive systematic search was conducted in the following electronic databases: PubMed/MEDLINE (1.02.2026), Cochrane Central Register of Controlled Trials/CENTRAL (2.02.2026), Web of Science[C9] (02.02.2026) and Scopus (8.03.2026).

b) Search strategy: An extensive literature search was conducted to identify studies evaluating the predictive and diagnostic value of sCysC for AKI in neonates. The search strategy was developed using three core concepts combined with Boolean operators: (1) neonatal population terms, (2) sCysC biomarker terms, and (3) AKI. Data specific syntax and controlled vocabulary were applied for each involved database, including Medical Subject Headings (MeSH).

Complete search strategies for all databases were provided in Supplementary Material 3.c) Limits and Restrictions: During the initial search no language or publication date restrictions were applied. Furthermore, no study design filters were applied.

D. Study Selection and Data Extraction:

All records retrieved from the database searches were imported into Zotero for de-duplication. Study selection was conducted in two sequential stages: (1) title and abstract screening, followed by (2) full-text eligibility assessment.

The data extraction after the screening was completed using a standardized extraction form, which was initially tested on a subset of studies to ensure consistency. Extracted data included study characteristics, population demographics, assay methods for sCysC, AKI definitions, timing of biomarker measurement, and diagnostic accuracy outcomes, including true positives, false positives, false negatives, true negatives, sensitivity, specificity, and AUC-ROC.

Both, screening stages and data extraction were performed independently by two reviewers using Rayyan, with reviewers blinded to each other’s decisions. Disagreements were resolved through discussion with a third reviewer.

E. Relationship between index test and reference standard according to timing:

The selected studies were grouped into two categories based on the timing between sCysC measurement (index test) and AKI diagnosis using sCr (reference standard):

-

Concurrent diagnostic accuracy studies: sCysC and sCr were measured simultaneously or within 24 hours, enabling direct comparison of the biomarkers. This design assessed the diagnostic accuracy of sCysC in detecting AKI, concurrently with the reference standard.

-

Predictive accuracy studies: sCysC was measured more than 24 hours before AKI was diagnosed using creatinine-based criteria. This design evaluated the ability of sCysC to identify newborns at risk of developing AKI before traditional biomarkers had abnormal levels.

This time-related classification highlighted the dual clinical utility of sCysC: earlier measurement assessed its value as a predictive biomarker, while concurrent measurement assessed its value as a diagnostic method. The 24-hour window was chosen based on previous studies that demonstrated that sCysC increases 1-2 days before serum creatinine in AKI among neonates.

F. Risk of Bias Assessment

The risk of bias of the included studies was evaluated by two independent assessors using the Quality Assessment of Diagnostic Accuracy studies 2 (QUADAS-2) tool, with signalling questions for the neonatal AKI topic. Any disagreements between the reviewers were resolved through discussion and consensus, or through discussion with a third reviewer. In particular, studies employing case-control designs that included completely healthy neonates as controls were flagged as having potential spectrum bias. Additionally, studies were evaluated for potential treatment-informed bias, defined as situations where sCysC results were available to clinicians and may have influenced clinical management, potentially affecting subsequent creatinine-based AKI diagnosis.

G. Data synthesis and analysis

Previous literature demonstrated significant heterogeneity, which prevented meta-analysis. This systematic review also identified substantial clinical and methodological heterogeneity among included studies. Key sources included variability in AKI definitions, differences in sCysC assessment and cut-off values, diverse study populations, and inconsistent timing between index test measurement and reference standard assessment. Following the SwiM reporting guideline, we described heterogeneity narratively rather than using statistical pooling. Reported diagnostic performance varied considerably; in the predictive group, AUC values ranged from 0.670 to 1.00, reflecting differences in population, clinical context, AKI definition, and measurement timing that could not be addressed through statistical pooling. Data were synthesized using a structured narrative synthesis approach in accordance with the Synthesis Without Meta-analysis (SWiM) reporting guideline. To address the primary research question and to preserve the distinction between clinical utilities, the studies evaluating concurrent diagnostic accuracy and those who assessed early predictive accuracy were analyzed separately.

Diagnostic performance metrics were summarized and they included sensitivity, specificity, positive and negative predictive values, and area under the receiver operating characteristic curve (AUC).

Sensitivity and Subgroup Analyses

To explore heterogeneity and assess the robustness of findings, the following narrative sensitivity analyses had been pre-specified: (1) exclusion of case-control studies to evaluate the impact of spectrum bias on diagnostic accuracy estimates; (2) stratification by sCysC assay method (e.g PENIA, ELISA, PETIA) to account for calibration variances; and (3) exclusion of studies with small sample size (n<30) to assess the influence of imprecise estimates on overall findings. Furthermore, predefined subgroup analysis was established to evaluate the diagnostic accuracy of sCysC separately for preterm and term neonates, because of the different degrees of renal maturation.

In addition, there were no deviations from the prospectively registered PROSPERO protocol occurring during the conduct of this systematic review.

3. Results

3.1. Study Selection

We conducted a systematic literature search using the selected electronic database and we identified a total of 568 records (PubMed 105 records, Web of Science 139 records, Cochrane Central 10 records and Scopus 314 records) (Figure 1). After removal of 203 duplicate records, 366 unique records remained for text and abstract screening.

During the initial screening phase, 346 studies were excluded based on the eligibility criteria: mixed pediatric population, absence of sCysC measurement, and no diagnostic study designs (e.g., reviews, meta-analysis, editorials, or study on animals). Therefore, we assessed the full text screening of the remaining 19 articles and 5 articles were excluded for specific reasons. Those reasons are detailed in the Supplementary Materials Table S1. The main reason of exclusion was a lack of mandatory diagnostic accuracy metrics.

A total of 15 studies met all inclusion criteria and were included in the final synthesis. The study selection process is illustrated in the PRISMA-DTA flow diagram (Figure 1).

3.2. Characteristics

Fifteen studies met the inclusion criteria. They were published between 2012 and 2025, showing a rising interest in the new AKI biomarkers regarding the neonatal population. Studies were conducted across 6 countries, including Egypt (n=8 studies), China (n=3 studies), Turkey (n=1 study), Greece (n=1 study), Slovenia (n=1 study), and Indonesia (n=1 study). The majority of study designs were prospective case-control (n=8 studies); one study had a mixed prospective case-control and cohort design. The rest of the studies employed either prospective cohorts (n=3 studies) or observational design (n=1 study), or retrospective (n=2 studies) cohorts designs.

Population. The included studies comprised a total of 53,751 neonates. Sample sizes ranged from 30 to 52,333 neonates (a median number of 90). Relating to gestational age, 8 studies enrolled only term neonates, 4 studies included only preterm neonates, and 3 studies included mixed populations (both term and preterm neonates). Clinical context varied across studies, with the most common conditions being perinatal asphyxia or HIE (n=5 studies), mixed critically conditions including sepsis (n=5 studies), RDS (n=4 studies), and severe hyperbilirubinemia (n=1 study).

Index test. The type of methods used for measuring sCysC across the included studies were the following: immunosorbent assay (ELISA) in 8 studies, particle-enhanced nephelometric immunoassay (PENIA) in 3 studies, particle enhanced turbidimetric immunoassay (PETIA) in 2 studies, immunonephelometry in 1 study and unspecified validated laboratory assay in 1 study. The timing of sCysC measurement varied, ranging from 0 hours (cord blood) to 72 hours of life. The threshold values ranged from 1.25 mg/L to 9.40 mg/L.

Reference standard. KDIGO-adapted neonatal criteria was used in 5 studies to define AKI, modified KDIGO criteria in 3 studies, pRIFLE/nRIFLE/AKIN criteria in 3 studies, and institution-specific definition or absolute sCr thresholds in 4 studies. The incidence of AKI across the included studies ranged from 12.1% to 61.5%.

Classification based on sampling timing. Based on timing between sCysC measurement and AKI diagnosis, 4 studies evaluated concurrent diagnostic accuracy (sCysC measured within the same 24 hours window as AKI diagnosis), 12 evaluated early predictive accuracy (sCysC measured ≥24 hours before AKI diagnosis) with 1 study [21] contributing data to both analyses due to its mixed timing design.

The detailed characteristics of all included studies are summarized in Table 1.

3.3. Risk of Bias and Diagnostic Accuracy

Risk of bias assessment. All 15 included studies were assessed for methodological quality using the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2). Results are summarized in Figure 2.

3.3.1. Risk of Bias Domains

Patient selection. The risk of bias in patient selection was low in 4 studies and high in 11 studies. The main sources of bias were the use of case-control study designs comparing sick neonates with healthy controls, alongside non-consecutive patient enrollment, which collectively may lead to spectrum bias.

Index test. The risk of bias in the index test domain was high in 13 studies, low in 1 study, and unclear in 1 study. A notable concern was that sCysC thresholds were predominantly determined post-hoc using ROC curve analysis rather than testing a pre-defined cut-off value. Therefore, none of the included studies explicitly reported blinding of index test interpretation to reference standard results.

Reference standard. Regarding the reference standard, the risk of bias was low in 7 studies, high in 3 studies, and unclear in 5 studies. The use of sCr-based criteria is an imperfect gold standard in neonates due to delayed response to renal injury, maternal creatinine transfer and gestational age dependence. This limitation may underestimate the true diagnostic accuracy of sCysC.

Flow and Timing. Most of the studies showed appropriate time intervals between sCysC measurement and sCr assessment. There was one study, [23], that was rated as high risk of bias due to a small number of AKI cases (n=6) which may have affected the diagnostic accuracy estimates. Concerns in 2 other studies - Xu et al. [13], Hidayati et al. [27]- included variable timing of sample across the neonatal period.

3.3.2. Applicability Concerns

Patient selection. Applicability concerns were rated low across all studies for patient selection, as the population accurately reflected the targeted neonatal demographic.

Index test. Applicability concerns for index tests across all the included studies, reflecting that the laboratory assays for sCysC, were validated.

Reference standard. High or unclear concerns regarding reference standard applicability was noted in 7 studies due to non standardized AKI definitions or exclusion of urine output criteria from the diagnostic method.

3.3.3. Overall Assessment

Regarding overall risk, no single study was rated to have a low risk of bias across all four domains. Fourteen studies (93%) had high risk in at least one domain, while one study (Xu et al. [12]) presented a moderate overall risk of bias.

The high prevalence of methodological concerns across the included studies reflects three structural problems in the existing literature. First, spectrum bias appears in several studies that used case-control designs comparing critically ill neonates with healthy controls rather than with high-risk neonates without AKI. Second, post hoc threshold selection is common: 13 of the 15 studies derived the optimal sCysC value by analyzing their own outcome data rather than testing a predefined cut-off value, a practice that systematically overestimates performance. Third, and perhaps the most important problem, concerns the reference standard against which sCysC performance was measured. Serum creatinine, universally used across the included studies, rises substantially later than sCysC. Consequently, an early elevation of sCysC that precedes creatinine-defined AKI is recorded as a false positive rather than recognized as an early predictive signal.

3.4. Sensitivity and Subgroup Analyses

As specified in the predefined review protocol, a narrative synthesis was performed, including sensitivity analysis and subgroup analysis to address the existing methodological heterogeneity.

3.4.1. Sensitivity Analyses

To address the potential spectrum bias, a sensitivity analysis was performed by excluding case-control studies. Following the exclusion of these studies, 6 prospective or retrospective cohort studies remained, and the predictive value of sCysC was found to remain consistently high. The consistently high AUC values, ranging from 0.84 to 0.93, indicate that the conclusions remained robust and that the diagnostic performance of sCysC was not artificially inflated by the case-control design studies present in the literature.

In a secondary analysis, the extracted data were stratified according to the laboratory assay methodology used for sCysC measurement. ELISA assay (n=8) reported strong diagnostic performance, with AUC values ranging from 0.73 to 1.00, Sarafidis et al.[25] reporting an AUC of 0.73 and Diab et al.[29] showed AUC of 1.00). Comparable results with ELISA-based studies were noted in studies utilizing automated particle-enhanced approaches (PENIA, n=3, PETIA, n=2) which AUC values were between 0.67 (Zhang et al. [31]) to 0.97 (Abdelaal et al. [22]). These findings support that the diagnostic accuracy of sCysC is assay–independent across different population, provided that a validated calibration protocol is followed.

To assess the reliability of the findings, a sensitivity analysis excluded studies with fewer than 30 neonates. All included studies met this criterion, with the smallest cohorts comprising 30 neonates, as reported by Nour et al. [24] and El-Frargy et al. [26]. Therefore, no studies were excluded. The robustness of the synthesis is further supported by the substantial sample sizes within the cohort subset, particularly the 52,333 neonates analyzed by Xu et al. [13].

3.4.2. Subgroup Analysis: Term vs Preterm Neonates

To assess whether gestational age may modify the early predictive and diagnostic utility of sCysC, a predefined subgroup analysis comparing preterm and full-term neonates was conducted. Within the preterm subgroup, where RDS represented the predominant underlying condition – as evidenced by the studies of Abdelaal et al. [22], El-Gammacy et al. [10], Elmas et al. [23] and Afifi et al. [29] – sCysC demonstrated excellent early predictive accuracy, with AUC values ranging from 0.88 to 1.00. Among full term neonates, predominantly presenting with perinatal asphyxia or HIE – as documented by Sarafidis et al. [25], Treiber et al. [28], Refat et al. [2] – sCysC demonstrated strong discriminatory performance for AKI, with AUC values approaching 0.939.

Interventionary studies involving animals or humans, and other studies that require ethical approval, must list the authority that provided approval and the corresponding ethical approval code.

4. Diagnostic Accuracy Findings

Of the 15 studies, three were classified as concurrent, with sCysC measured within 24 hours of AKI diagnosis. Eleven were early predictive, with sCysC measured more than 24 hours before diagnosis. Abdelsattar et al. [21] was classified as both. Table 2 summarizes the diagnostic performance of all studies by assessment timing.

a)

Concurrent Diagnostic Accuracy:

Sensitivity and specificity. Analyses showed that sCys Csensitivity values ranged from 84,8% to 88,5%, while specificity ranged from 61.8% to 75%. The highest sensitivity (88.46%) and specificity (75%) were reported by Abdelsattar et al. [21] in neonates with sepsis associated AKI. Xu et al. [13], did not report sCysC sensitivity or specificity values.

Area under the ROC. Of the four concurrent diagnostic accuracy studies, two reported AUC values ranging from 0,844 to 0,849. All four studies demonstrated good discriminatory ability (AUC values ≥0.80) for sCysC. The highest AUC (0.849) was reported by Hidayati et al. [27] in critically ill newborns admitted in the NICU.

Optimal threshold values. Reported sCysC thresholds reflected a substantial variability. There were three thresholds ranging from 1.55 to 2.2 mg/L, consistent with published reference values. One study (Abdelsattar et al. [21]) reported a higher sCysCthreshold of >9.4 mg/L in newborns with sepsis-associated AKI. This significantly higher threshold might reflect the use of a non-standardized ELISA assay method and a heterogeneous neonatal population with severe septic shock.

b)

Early predictive diagnostic accuracy

Sensitivity and specificity. Serum cystatin C sensitivity varied substantially across the 12 early predictive studies, ranging from 16% to 100% (Table 2), that reflects methodological heterogeneity as much as true variation in biomarker performance. The lowest value in this range – 16%, reported by Elmas et al. [23] - in preterm neonates with RDS, using a cut off of ≥ 1.62 mg/L, derives from a study in which only 6 neonates developed AKI, a sample size that is insufficient to support a reliable sCysC sensitivity estimation and this result should not be read as representative of how the biomarker performs. Excluding it, sCysC sensitivity across the remaining studies ranged from 53.3% to 100%.

Two studies at the upper end of this range – Abdelaal et al. [22] and Afifi et al. [29], both reporting 100% sensitivity – used case-control designs that compared RDS neonates with healthy controls, a design known to introduce spectrum bias and to overestimate discriminatory performance.

Among studies less susceptible to spectrum bias, sCysC sensitivity ranged from 61.5% in asphyxiated neonates to 95% (Refat et al. [2], in a prospective cohort with perinatal asphyxia).

The reported sCysC specificity ranged from 69.3% to 100% across the same studies. Perfect specificity (100%) was reported by El-Sadek et al. [20] and Afifi et al. [29]; both used case-control designs incorporating healthy controls, a methodological choice that makes it substantially easier for any biomarker to achieve high specificity by comparing fundamentally different populations. In contrast, the lowest sCysCspecificity (69.3%), reported by Zhang et al. [31], was observed in a study where the comparison group consisted of asphyxiated neonates without AKI – a design that more accurately reflects the clinical setting in which the biomarker would be applied.

Area under the ROC. The AUC values were reported by 10 of the 12 early predictive studies, ranging from 0.670 to 1.00. Eight studies indicated good to excellent discriminatory ability (with ≥0.80). The lowest sCysCAUC (0.670) was reported by Zhang et al. [31] in asphyxiated neonates. Two studies did not reported AUC data (Nour et al. [24], El-Frargyet al. [26]).

Diagnostic lead time. Analysis showed that the time interval by which sCysC elevation preceded sCr-defined AKI ranged from 24 hours to approximately 4 days. The diagnostic lead time reported by most of the studies was 24 to 72 hours, with the longest interval (approx. 4 days) reported by Abdelaal et al. [22] in preterm neonates with RDS.