Review Reports - Acute Kidney Injury in Critically Ill Children: A Retrospective Single-Center Study

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

I have reviewed with interest this manuscript titled "Acute Kidney Injury in Critically Ill Children: A Retrospective Single-Center Study." The study addresses a clinically relevant question, namely the epidemiology and outcomes of pediatric AKI in a resource-defined setting, and is of genuine interest to the clinical nephrologist, as it contributes regional MENA data to a literature still dominated by the large North American and European cohorts that have defined the field, most notably the multinational AWARE study (Kaddourah A, Basu RK, Bagshaw SM, Goldstein SL; AWARE Investigators. N Engl J Med 2017;376:11–20. https://doi.org/10.1056/NEJMoa1611391), against which the present incidence estimate should be read.

The methods are broadly appropriate for a retrospective single-center cohort, employing standardized KDIGO criteria, though several design choices require clarification and one carries a real risk of misclassification. The results are largely coherent with internal logic and with established pathophysiology: the gradient between AKI severity and adverse outcomes is consistent and biologically plausible, and the dissociation between AKI stage and mortality in the multivariable model (favoring inotrope requirement and complication burden) is a credible and well-reasoned finding. The discussion is reasonably balanced and appropriately tempers the RRT "care gap" interpretation with alternative explanations, though the conclusions occasionally outpace what a single-center retrospective design can support.

I would also encourage the authors to position this work within a broader conceptual frame in the Discussion. The methodological architecture they employ, namely a standardized diagnostic/staging definition (KDIGO) applied to a retrospective cohort, with severity stratification linked to outcomes and multivariable adjustment for confounders, is not specific to AKI. It is a transferable template that has proven informative across other clinical entities in glomerular and tubulointerstitial nephrology, where disease classification, severity gradation, and outcome prediction are similarly central. The same logic underlies the way pathogenesis-based stratification has reshaped the management of primary membranous nephropathy, where serological and risk-based classification now guides therapy and prognosis, the analogous histopathological-to-mechanistic framing applied to FSGS and the podocytopathy spectrum, or MGRS. A brief sentence acknowledging that the analytic approach and its limitations, particularly around definition-dependent classification and severity-outcome inference, generalizes to these other entities would strengthen the manuscript's framing and broaden its readership beyond the pediatric critical care community.

Below I detail my concerns:

1) the manuscript states that for patients without a prior creatinine value, baseline was "estimated using the Schwartz formula based on age-specific normal GFR values." This is a critical methodological vulnerability. Back-calculating baseline creatinine by assuming a normal GFR is known to inflate AKI incidence and severity, because any patient with unrecognized antecedent renal impairment is assigned an artificially low baseline. The authors must report what proportion of the 162 AKI patients had a true measured baseline versus an estimated one, and ideally provide a sensitivity analysis restricted to patients with measured baseline creatinine. Without this, the central incidence figure (22.5%) and the stage distribution rest on an unquantified assumption.

2) the text states 1,313 admissions were screened and 1,200 were eligible after exclusions (Section 3.1), but Figure 1 lists 113 excluded for reasons that include "Insufficient data," a category not described in the Methods exclusion criteria (Section 2.2). If patients were excluded for insufficient data, this is a source of selection bias that must be stated explicitly in the Methods and quantified. The Methods and Figure 1 must be reconciled, and any data-completeness exclusion justified.

3) the stage 3 subgroup comprises 54 patients, of whom only 6 (11.1%) received RRT and 15 (27.8%) died. The manuscript repeatedly frames the low RRT utilization (4.3% overall) as a potential "care gap" that "may have contributed to preventable mortality." This inference is not supported by the data presented, which contain no information on whether non-dialyzed Stage 3 patients met institutional RRT criteria. The claim should be softened to a hypothesis-generating observation.

4) the binary logistic regression for mortality was built on only 27 events. With forward stepwise selection and the number of candidate predictors described (age, sex, comorbidities, inotropes, MV duration, complications, admission type, LOS), the events-per-variable ratio is well below the conventional minimum of ~10, raising a substantial risk of overfitting and unstable estimates. This is consistent with the implausibly wide confidence interval for inotropic support (OR 6.12, 95% CI 1.87–22.97). The authors should state the final number of variables retained, justify the variable-selection strategy (forward stepwise selection is itself prone to bias), and acknowledge the events-per-variable limitation explicitly.

5) figure 5: it reports "AKI Stage 3 vs. Stage 1" as a statistically significant predictor of mortality (OR 3.18, p=0.016), yet the text (Section 3.4 and Discussion) states that AKI stage did not independently predict mortality once inotropes and complications were included. These two statements are contradictory as presented.

6) the variable “number of clinical complications” is one of the two strongest mortality predictors (OR 4.22) and a predictor of severity, yet it is never operationally defined. What counts as a complication?

Addressing these areas will substantially strengthen the manuscript's scientific rigor and the defensibility of its central claims.

Author Response

Reviewer provided a thorough and expert evaluation. We are grateful for the detailed, constructive critique and have addressed all specific concerns below.

Reviewer Comment R1-C1: The manuscript states that for patients without a prior creatinine value, baseline was 'estimated using the Schwartz formula based on age-specific normal GFR values.' This is a critical methodological vulnerability. Back-calculating baseline creatinine by assuming a normal GFR is known to inflate AKI incidence and severity. The authors must report what proportion of the 162 AKI patients had a true measured baseline versus an estimated one, and ideally provide a sensitivity analysis restricted to patients with measured baseline creatinine.

Response: We fully acknowledge this as an important limitation and thank the reviewer for raising it explicitly. The use of the Schwartz formula for estimating baseline creatinine in patients without prior measurements is a widely accepted, standard approach in pediatric retrospective AKI research (e.g., the AWARE study cohort, Kaddourah et al., NEJM 2017), precisely because historical creatinine values are systematically absent for many children who are otherwise healthy prior to acute illness. In this retrospective cohort, the exact proportion of patients relying on Schwartz-estimated versus laboratory-measured baseline creatinine was not prospectively tracked as a separate variable, and individual-level reconstruction of this count from archival records was not feasible within the scope of the revision. However, we have added an explicit acknowledgment of this limitation, including the known direction of potential bias (upward misclassification of AKI incidence and severity) to the Limitations section of the revised manuscript. We have also added a recommendation that future prospective studies document the proportion of estimated versus measured baselines to allow sensitivity analyses. [Change highlighted in yellow in the revised manuscript "Limitations" paragraph, Section 4.]

Reviewer Comment R1-C2: The text states 1,313 admissions were screened and 1,200 were eligible after exclusions (Section 3.1), but Figure 1 lists 113 excluded for reasons that include 'Insufficient data,' a category not described in the Methods exclusion criteria (Section 2.2). This is a source of selection bias that must be stated explicitly in the Methods and quantified. The Methods and Figure 1 must be reconciled.

Response: The reviewer is correct. 'Insufficient data' was indeed an operational exclusion applied during data extraction (specifically, patients with absent or unverifiable key variables such as missing baseline creatinine with no prior records and no identifiable age-based estimate, missing urine output documentation, or incomplete admission records), but this criterion was inadvertently omitted from the written Methods section. We have revised Section 2.2 (Patient Selection) to explicitly list 'patients with insufficient data to determine AKI status (e.g., absence of baseline creatinine and no verifiable prior records, or missing urine output data for the critical admission period)' as an additional exclusion criterion, and we have also revisied the Figure 1. Change highlighted in yellow Section 2.2.

Reviewer Comment R1-C3: The Stage 3 subgroup comprises 54 patients, of whom only 6 (11.1%) received RRT and 15 (27.8%) died. The manuscript repeatedly frames the low RRT utilization (4.3% overall) as a potential 'care gap' that 'may have contributed to preventable mortality.' This inference is not supported by the data presented, which contain no information on whether non-dialyzed Stage 3 patients met institutional RRT criteria. The claim should be softened to a hypothesis-generating observation.

Response: We fully accept this critique. The phrase 'care gap' and the associated language implying that low RRT utilization 'may have contributed to preventable mortality' goes beyond what our retrospective data can demonstrate, since we have no information on whether non-dialyzed Stage 3 patients met institutional criteria for RRT initiation. We have revised the relevant passages in the Discussion section to reframe this as a hypothesis-generating observation warranting prospective study, removing the causal language. The revised text now reads: 'The low RRT utilization rate observed in this cohort raises a hypothesis-generating question about whether standardized nephrology referral protocols and clearer institutional RRT initiation thresholds might influence outcomes; prospective studies with granular fluid overload and decision-audit data are needed to evaluate this.' [Change highlighted in yellow "Discussion" section, RRT paragraph, in the revised manuscript.]

Reviewer Comment R1-C4: The binary logistic regression for mortality was built on only 27 events. With forward stepwise selection and the number of candidate predictors described, the events-per-variable (EPV) ratio is well below the conventional minimum of ~10, raising a substantial risk of overfitting and unstable estimates. This is consistent with the implausibly wide confidence interval for inotropic support (OR 6.12, 95% CI 1.87–22.97). The authors should state the final number of variables retained, justify the variable-selection strategy, and acknowledge the EPV limitation.

Response: We fully accept this statistical concern and have substantially revised the regression approach in the resubmission. Rather than forward stepwise selection, the revised model uses a simultaneous-entry design: all four candidate predictors — gender, inotropic support, number of clinical complications, and PRISM IV score — were entered simultaneously in the multivariable model, and each was also evaluated in a separate unadjusted model. This directly addresses the stepwise variable-selection bias concern. The revised multivariable results are: inotropic support (adjusted OR 8.51, 95% CI: 2.44–29.63; p < 0.001), number of complications (adjusted OR 4.38, 95% CI: 2.28–8.42; p < 0.001), PRISM IV score (adjusted OR 1.07, 95% CI: 1.00–1.14; p = 0.042), and gender (adjusted OR 2.24, 95% CI: 0.63–7.92; p = 0.212, not significant). We acknowledge that with 27 mortality events and four predictor variables the events-per-variable (EPV) ratio is 6.75, which remains below the conventional threshold of 10. We have retained the EPV limitation statement in the revised Limitations section and recommend that future larger prospective studies confirm these associations with adequate EPV. [Changes highlighted in yellow Section 2.5 and Results §3.4 in the revised manuscript.]. Further Variable gender is included in the multivariable model for adjustment purposes, if excluded from the model the overall significance of predictors remain same and the EPV ratio can be improved form 6.75 to 9, which is very close to conventionla threshold of 10.

Reviewer Comment R1-C5: Figure 5 reports 'AKI Stage 3 vs. Stage 1' as a statistically significant predictor of mortality (OR 3.18, p=0.016), yet the text (Section 3.4 and Discussion) states that AKI stage did not independently predict mortality once inotropes and complications were included. These two statements are contradictory as presented.

Response: The reviewer has identified a genuine inconsistency between the forest plot and the narrative text that we apologize for and have now corrected. Based on the comments of other reviewer, we have replaced forest plot with Table 3 and added missing table 2. [Changes highlighted in yellow Section 3.4 in the revised manuscript.]

Reviewer Comment R1-C6: The variable 'number of clinical complications' is one of the two strongest mortality predictors (OR 4.38) and a predictor of severity, yet it is never operationally defined. What counts as a complication?

Response: This is an important omission that we have corrected. We have added an operational definition to Section 2.4 (Data Collection). Clinical complications were defined as discrete, clinically documented adverse events occurring during the PICU stay, identified from nursing and physician notes. Counted complications included: (1) fluid overload (≥10% above baseline body weight); (2) metabolic acidosis requiring intervention (pH < 7.2); (3) hyperkalemia requiring active management (serum K⁺ > 6.0 mEq/L); (4) hypocalcemia requiring supplementation; (5) pulmonary edema or new respiratory deterioration attributed to fluid overload; (6) new-onset arrhythmia; and (7) any additional organ system failure (beyond the primary presenting diagnosis) documented by the treating intensivist. Each distinct event was counted separately, yielding an integer count per patient. [Change highlighted in yellow Section 2.4 in the revised manuscript.]

Reviewer 2 Report

Comments and Suggestions for Authors

The authors exemined the incidence and severity of AKI in children who cared in PICU in their institution. They found that Early AKI recognition using KDIGO criteria, heightened vigilance in younger children, and prompt hemodynamic stabilization are essential strategies to reduce morbidity and mortality in critically ill pediatric patients. This MS is potentially interesting. However, novel information remains limited.

This theme is not new, somewhat second hand topics. Thus, some novelty worth publication should be represented clearly.
Looking at the MS, recent papers regarding this theme are lacking, i.e.,
Can J Kidney Health Dis. 2023 Jun 12;10:20543581231168088. doi: 10.1177/20543581231168088.

EClinicalMedicine. 2024 Jan 5;68:102409. doi: 10.1016/j.eclinm.2023.102409.
It is necessary to dsicuss this theme in more depth using such recent papers listed above.
It is necessary to discuss the similarities and differences of the main findings between the authors' study and previously reported studies regarding the same theme.
The limitationsof this study should be stated cleraly.
What kind of differences existed in the results obtained between the authors' country and the other high economic state countries ?
The whole description should be rewritten cocisely and logically.

Comments on the Quality of English Language

Clear, concise and logical writing style should be encouragged.

Author Response

Reviewer found the manuscript 'potentially interesting' and raised seven comments relating to novelty, recent literature coverage, depth of comparative discussion, and clarity. We are grateful for the detailed, constructive critique and have addressed all specific concerns below.

Reviewer Comment R2-C1: This theme is not new, somewhat second-hand topics. Thus, some novelty worth publication should be represented clearly.

Response: We respectfully note that while AKI in pediatric critical care is an established research area, this study makes several distinct contributions that justify publication and that we have now made more explicit in the revised Introduction: This study makes four distinct contributions to the existing literature: (1) it provides the first five-year longitudinal epidemiological data from a Saudi Arabian tertiary PICU using the full KDIGO 2012 criteria; (2) it describes a regional comorbidity pattern, with predominance of hematology-oncology cases, which differs from the cardiac sur-gery-dominant Western cohorts that have historically shaped AKI risk stratification; (3) it shows that mortality was more strongly associated with hemodynamic instability and overall complication burden than with AKI stage alone; and (4) it documents a low RRT utilization rate of 4.3% despite a substantial proportion of Stage 3 AKI, raising an im-portant question for future research and PICU policy in resource-limited settings. These points have been added to the final paragraph of the Introduction. [Change highlighted in yellow "Introduction", final paragraph.]

Reviewer Comment R2-C2/3: Looking at the MS, recent papers regarding this theme are lacking.' The reviewer cites: Can J Kidney Health Dis. 2023 Jun 12;10:20543581231168088 and EClinicalMedicine. 2024 Jan 5;68:102409. 'It is necessary to discuss this theme in more depth using such recent papers listed above.

Response: We thank the reviewer for directing us to these recent publications, which we have now incorporated into the Discussion. Both papers were reviewed and cited as follows: The Can J Kidney Health Dis 2023 paper (DOI: 10.1177/20543581231168088) provides updated epidemiological data on pediatric AKI outcomes in a North American setting; we have cited it when contextualizing our mortality rates against contemporary benchmarks, noting points of convergence (KDIGO staging as a prognostic tool) and divergence (lower RRT utilization in our MENA cohort). The EClinicalMedicine 2024 paper (DOI: 10.1016/j.eclinm.2023.102409) offers relevant multi-center data on AKI risk factors in critically ill children; we have incorporated a comparison of its key predictor variables against our findings in the Discussion, noting the consistent role of hemodynamic instability across settings. Both references have been added to the reference list. [Changes highlighted in yellow "Discussion" paragraphs 3-4 and reference list in the revised manuscript.]

Reviewer Comment R2-C4: It is necessary to discuss the similarities and differences of the main findings between the authors' study and previously reported studies regarding the same theme.

Response: The Discussion already includes comparison with several key studies (AWARE study, Alkandari et al., Al-Otaibi et al., Kari et al., regional MENA studies). In the revision, we have expanded this comparative discussion more systematically, structuring a dedicated paragraph that explicitly contrasts our incidence (22.5%), mortality (16.7%), and RRT utilization (4.3%) against the AWARE study (26.9% incidence, ~20% mortality), the two newly added papers (Comments 2–3 above), and regional MENA data. We also highlight the unique hematology-oncology comorbidity profile of our cohort versus the cardiac surgery-dominant North American cohorts. [Changes highlighted in yellow "Discussion" comparative paragraph in the revised manuscript.]

Reviewer Comment R2-C5: The limitations of this study should be stated clearly.

Response: The Limitations paragraph in the original Discussion has been substantially expanded. In addition to the existing limitations (retrospective single-center design, Schwartz formula baseline estimation, incomplete UOP documentation, no long-term follow-up), we have added: (1) the EPV limitation in the mortality regression (per R1-C4); (2) the unquantified proportion of Schwartz-estimated baselines (per R1-C1); (3) the omission of PRISM IV from the multivariable model (per R3-C2); (4) the clinical and physiological heterogeneity inherent in an age range of 29 days to 14 years (per R3-C5); and (5) the lack of granular nephrotoxin exposure and fluid balance data. [Changes highlighted in yellow "Limitations" paragraph in the revised manuscript.]

Reviewer Comment R2-C6: What kind of differences existed in the results obtained between the authors' country and the other high economic state countries?

Response: We have added an explicit comparative paragraph in the Discussion addressing this directly. Key differences identified: (1) Incidence: our 22.5% is modestly lower than the 26.9% in the multinational AWARE study, likely reflecting our strict exclusion of post-operative elective admissions and neonates. (2) RRT utilization: 4.3% in our cohort versus 20–30% in high-income country reports, reflecting resource constraints, conservative decision thresholds, and the absence of a standardized institutional RRT protocol. (3) Comorbidity profile: hematology-oncology predominance in our cohort versus cardiac surgery-related AKI in North American and European PICUs. (4) Mortality: our overall 16.7% is comparable to high-income country benchmarks (10–20%), but the Stage 3 mortality (27.8%) approaches the higher rates seen in some resource-limited settings, likely reflecting the absence of early CRRT. [Changes highlighted in yellow "Discussion" comparative paragraph in the revised manuscript.]

Reviewer Comment R2-C7: The whole description should be rewritten concisely and logically.

Response: We respectfully note that this comment is not accompanied by specific examples of sections or passages that are unclear or redundant. Reviewer 1 assessed the manuscript as 'broadly appropriate' with a discussion that is 'reasonably balanced,' and Reviewer 3 described the study as 'well-written.' We have, however, taken this comment as encouragement to tighten the narrative where possible: the Introduction's final paragraph has been condensed, several transitional sentences in the Discussion have been streamlined, and the Conclusions section has been made more concise. We believe these targeted edits, together with the substantive revisions made in response to all other comments, materially improve the manuscript's clarity and logical flow without a wholesale rewrite that would risk disrupting the paper's existing coherent structure. [Minor edits throughout general tightening, not individually highlighted.]

Reviewer 3 Report

Comments and Suggestions for Authors

First of all, I would like to congratulate the authors on a well-written and clinically relevant study. The investigation of Acute Kidney Injury (AKI) in critically ill children is of paramount importance, and this 5-year retrospective cohort from a tertiary Pediatric Intensive Care Unit in Saudi Arabia provides a very valuable addition to the literature.

However, to maximize the scientific rigor and the validity of the final conclusions, I suggest addressing a few methodological and statistical aspects before publication.

Regarding the statistical analysis, there is a significant concern with how continuous variables are presented and tested. In the Methods section, it is correctly stated that non-normally distributed variables would be analyzed using the Kruskal-Wallis test. However, in Table 1 and throughout the text, all continuous variables are presented as Means with Standard Deviations (SD) and analyzed via ANOVA. Variables like Weight (Mean 17.6, SD 16.1 kg) and Age (Mean 4.7, SD 4.5 years) have a strict lower bound of zero. The fact that their standard deviations are almost as large as their means mathematically proves that these data follow a strongly right-skewed distribution, severely violating the normality assumption required for ANOVA. The authors should revisit their Shapiro-Wilk test results, present these skewed variables as Medians [IQR], and use the non-parametric Kruskal-Wallis test to compare stages, exactly as originally intended in the methodology.

Another crucial point relates to the multivariable logistic regression model used for PICU mortality. The text mentions that baseline illness severity, measured by the PRISM IV score, was comparable across all groups at admission (mean 11.0). However, it seems this score was not included as an adjusting covariate in the regression model. To confidently assert that the need for inotropes and the number of complications are truly independent predictors of mortality, and not merely surrogate markers for patients who were already sicker upon admission, the authors must run the multivariable model adjusting for the baseline PRISM IV score.

There are a few clarifications needed regarding missing data and clinical definitions. Figure 1 indicates that 113 patients were excluded due to "insufficient data." It would be helpful to briefly specify in the Methods what this entails (e.g., missing baseline creatinine, missing flowsheets).

Regarding data presentation, please consider including the baseline PRISM IV scores and the primary admission reasons (medical vs. surgical) directly as rows in Table 1. This would allow readers a quick visual comparison of the baseline characteristics. In Figures 2 and 3, I suggest moving the floating exact p-values to the figure legends or using standard asterisk notations on the bars for a cleaner visual look.

Finally, considering the massive variances observed in demographic data (with ages ranging from 29 days to 14 years), there is a profound clinical and physiological heterogeneity in the sample. The authors should briefly acknowledge this sample heterogeneity in the Limitations section, as it might obscure specific subgroup trends (e.g., infants versus older adolescents).

Author Response

We are grateful for this careful review and have addressed each point in full.

Reviewer Comment R3-C1: In Table 1, continuous variables (Age, Weight) have SD ≈ Mean, which mathematically demonstrates strongly right-skewed distributions, severely violating ANOVA's normality assumption. The authors should revisit Shapiro-Wilk results, present skewed variables as Medians [IQR], and use the non-parametric Kruskal-Wallis test — exactly as stated in the Methods.

Response: The reviewer is absolutely correct, and we thank them for this precise and valid statistical critique. The Methods section correctly specified that non-normally distributed variables would be presented as Medians [IQR] and compared using the Kruskal-Wallis test. On re-examination of our Shapiro-Wilk results, Age and Weight did not satisfy the normality assumption (both p < 0.05 on Shapiro-Wilk), which is further confirmed, as the reviewer notes, by the near-unit coefficient of variation (SD/mean ≈ 1.0 for both variables). Accordingly, Table 1 has been revised: Age and Weight are now presented as Median [IQR] and the between-stage comparisons for these variables now use the Kruskal-Wallis test. The p-values and effect estimates have been updated accordingly in both Table 1 and the corresponding in-text descriptions. Height was similarly evaluated and revised where appropriate. [Changes highlighted in yellow "Table 1" and Section 3.1 in the revised manuscript.]

Reviewer Comment R3-C2: The PRISM IV score was comparable across all admission groups (mean 11.0) but does not appear to have been included as an adjusting covariate in the regression model. To assert that inotropes and complications are truly independent predictors, the model must adjust for baseline PRISM IV score.

Response: We are pleased to report that this concern has been fully addressed in the revised model. The binary logistic regression has been re-estimated using a simultaneous-entry comparative approach that explicitly includes PRISM IV score as a pre-specified covariate alongside gender, inotropic support, and number of clinical complications. PRISM IV score is now a statistically significant independent predictor of PICU mortality in both the unadjusted model (OR 1.05, 95% CI: 1.01–1.09; p = 0.018) and the multivariable-adjusted model (OR 1.07, 95% CI: 1.00–1.14; p = 0.042), confirming that baseline illness severity independently predicts mortality even after controlling for hemodynamic instability and complication burden. The full revised regression table (Table 4) is included in the resubmission. [Changes highlighted in yellow "Section 2.5" and "Results 3.4" in the revised manuscript.]

Reviewer Comment R3-C3: Figure 1 indicates that 113 patients were excluded due to 'insufficient data.' It would be helpful to briefly specify in the Methods what this entails.

Response: This concern is shared with Reviewer 1 (R1-C2) and has been fully addressed there. Section 2.2 now explicitly defines 'insufficient data' as an exclusion criterion, and Figure 1 has been revised to itemize the reasons within this category. [Change highlighted in yellow Section 2.2 and Figure 1]

Reviewer Comment R3-C4: Please include baseline PRISM IV scores and primary admission reasons (medical vs. surgical) directly as rows in Table 1. In Figures 2 and 3, move floating exact p-values to figure legends or use standard asterisk notations for a cleaner visual look.

Response: We have implemented both suggestions. (1) Table 1 has been expanded to include PRISM IV score (Median [IQR] by AKI stage) and admission type (medical vs. surgical, n [%]) as additional rows, allowing readers direct visual comparison of baseline characteristics across stages. [Changes highlighted in yellow Table 1, in the revised manuscript.]

Reviewer Comment R3-C5: Considering the massive variances observed in demographic data (ages ranging from 29 days to 14 years), there is profound clinical and physiological heterogeneity. The authors should briefly acknowledge this in the Limitations section, as it might obscure specific subgroup trends (e.g., infants versus older adolescents).

Response: We fully agree and have added a dedicated sentence to the Limitations section explicitly acknowledging the wide age spectrum. The revised text reads: 'Furthermore, the study population spans a broad developmental and physiological continuum, from late neonates (29 days) to adolescents (14 years). This age heterogeneity may obscure age-specific subgroup trends for example, differences in AKI pathophysiology, clinical presentation, and outcomes between infants and school-age children that would only be detectable with stratified subgroup analyses in a larger prospective cohort.' [Change highlighted in yellow "Limitations" paragraph in the revised manuscript.]

Reviewer 4 Report

Comments and Suggestions for Authors

Dear authors,

The manuscript is clinically significant, and data from your region could be of interest. However, in its current form, the article has serious flaws that need to be addressed. First, there is a fundamental numerical inconsistency. You state that 162 out of 1,200 selected patients developed AKI, yet you report an incidence of 22.5%. This is incorrect; 162/1,200 is 13.5%. The correct denominator must be clarified immediately, and the abstract, results, flow diagram, and discussion must be corrected.

Moreover, it is not explained why the data cut-off is December 2022, given that the manuscript is being submitted now. You need to state whether there was a delay in analysis or writing, and whether PICU practice, the availability of nephrology consultations, or RRT protocols have changed since then.

I would also like to emphasize that, while the use of the KDIGO 2012 criteria is acceptable, the literature review and discussion need to be updated. You should comment on the more recent AKI/AKD framework and the importance of post-AKI follow-up, particularly given the absence of long-term renal outcome data.

Concerning the methodology, it lacks sufficient clarity. You need to specify how many patients were staged using creatinine criteria, how many using urine output criteria, and how many using both. Additionally, you should state how many had an actual baseline creatinine and how many had an estimated baseline calculated via the Schwartz formula. The statistical analysis is weak. The mortality model relies on only 27 deaths and four predictors, resulting in a low EPV. Consequently, statements regarding independent predictors should be phrased more cautiously. Lastly, the “number of complications” variable requires further explanation. Complications such as fluid overload, metabolic acidosis, hyperkalemia, and pulmonary edema may be consequences of AKI or indicators of severity rather than independent predictors. This issue needs to be addressed to avoid circular reasoning.

Finally, the manuscript requires significant linguistic and technical editing. There are errors such as “PRISMA IV” instead of “PRISM IV,” “Inotrops,” and “Complicaitons,” which undermine the credibility of the text. Overall, the topic is interesting, but the manuscript in its current form is not ready for publication. I recommend major revision.

Thank you.

Comments on the Quality of English Language

The English could be improved to more clearly express the research.

Author Response

We thank Reviewer for the careful evaluation of the manuscript and for identifying several issues that required correction or clarification. We have revised the manuscript to correct the numerical inconsistency in the AKI incidence, clarify the study-period cutoff, update the discussion of AKI recovery and post-AKI follow-up, refine the statistical interpretation, define the complications variable more explicitly, and correct technical and language errors. Changes in the revised manuscript are highlighted, with dataset-resolved methodological clarifications highlighted in green.

Reviewer Comment 1:

First, there is a fundamental numerical inconsistency. You state that 162 out of 1,200 selected patients developed AKI, yet you report an incidence of 22.5%. This is incorrect; 162/1,200 is 13.5%. The correct denominator must be clarified immediately, and the abstract, results, flow diagram, and discussion must be corrected.

Response: We agree with the reviewer. We corrected the AKI incidence throughout the manuscript to 13.5% among 1,200 eligible admissions. We also corrected the screened-admission percentage to 12.3% (162/1,313) and revised the abstract, highlights, Results, Discussion, Conclusions, and Figure 1 so the denominator and percentage are consistent.

Change made in the revised manuscript: Abstract; Highlights; Results, Section 3.1; Figure 1; Discussion; Conclusions.

Reviewer Comment 2:

Response: We have added a clarification that the December 2022 cutoff reflects the locked retrospective dataset used for IRB-approved extraction, data cleaning, statistical analysis, and manuscript preparation. We also added a limitation noting that any changes in PICU practice, nephrology consultation availability, or RRT protocols after December 2022 are not captured in this dataset.

Change made in the revised manuscript: Methods, Section 2.1; Discussion limitations.

Reviewer Comment 3:

Response: We agree that the manuscript needed clearer framing beyond the acute PICU episode. We added text describing AKI as part of a continuum that may progress to acute kidney disease and chronic kidney disease, and we emphasized that the absence of post-discharge renal follow-up limits assessment of renal recovery and longer-term outcomes.

Change made in the revised manuscript: Introduction; Discussion limitations; Conclusions.

Reviewer Comment 4:

Response: We thank the reviewer for pointing out this important methodological detail. We reviewed the available analysis dataset carefully. The dataset includes the final KDIGO stage for all 162 AKI patients (Stage 1, n = 44; Stage 2, n = 64; Stage 3, n = 54), admission urine output for all 162 patients, initial and highest creatinine values for all 162 patients, and GFR values for 160 patients. However, the dataset does not include separate audit fields showing whether each patient’s final KDIGO stage was determined by creatinine criteria alone, urine output criteria alone, or both criteria. It also does not include a separate variable indicating whether baseline creatinine was directly measured from prior records or estimated using the Schwartz formula. Because these classifications were not preserved as separate variables in the retrospective analysis file, we cannot reliably reconstruct those exact counts without risking misclassification. We have therefore revised the Methods section to clarify the available KDIGO-related variables and revised the limitations section to acknowledge that criterion-source and baseline-source counts could not be separately reported from the available dataset.

Change made in the revised manuscript: Methods, Section 2.3; Discussion limitations.

Reviewer Comment 5:

The statistical analysis is weak. The mortality model relies on only 27 deaths and four predictors, resulting in a low EPV. Consequently, statements regarding independent predictors should be phrased more cautiously.

Response: We agree and revised the interpretation throughout the manuscript. The statistical analysis now states the EPV ratio explicitly (27 deaths, four predictors, EPV 6.75), and the Results, Discussion, Highlights, Abstract, and Conclusions now use cautious language such as 'associated with mortality in the multivariable model' rather than presenting the variables as definitive independent causal predictors.

Change made in the revised manuscript: Highlights; Abstract; Statistical Analysis; Results, Section 3.4; Discussion; Conclusions.

Reviewer Comment 6:

Lastly, the 'number of complications' variable requires further explanation. Complications such as fluid overload, metabolic acidosis, hyperkalemia, and pulmonary edema may be consequences of AKI or indicators of severity rather than independent predictors. This issue needs to be addressed to avoid circular reasoning.

Response: We have expanded the operational definition of the complications variable and revised the interpretation to avoid circular reasoning. The manuscript now states that this variable should be interpreted as a marker of illness and AKI-severity burden rather than as a causal predictor independent of AKI pathophysiology.

Change made in the revised manuscript: Methods, Section 2.4; Results, Section 3.4; Discussion; limitations.

Reviewer Comment 7:

Finally, the manuscript requires significant linguistic and technical editing. There are errors such as 'PRISMA IV' instead of 'PRISM IV,' 'Inotrops,' and 'Complicaitons,' which undermine the credibility of the text.

Response: We corrected the technical wording and typographical errors identified by the reviewer, including PRISM IV terminology, 'inotropes/inotropic support,' 'complications,' 'Panel C,' and 'PICU length of stay.' We also made targeted language edits to reduce overstatement and improve clarity.

Change made in the revised manuscript: Tables, figure captions, Results, Discussion, and manuscript-wide terminology.

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

The revised MS has been much improved.

Comments on the Quality of English Language

May be OK.

Author Response

Dear Reviewer,

Thank you very much for your positive feedback. We are pleased that you found the revised manuscript much improved. We sincerely appreciate your thoughtful comments and constructive guidance, which helped us strengthen the quality and clarity of the manuscript.

Reviewer 3 Report

Comments and Suggestions for Authors

Dear Authors,

Thank you for your thorough and diligent revisions. The manuscript has significantly improved, particularly regarding the statistical rigor of the multivariable model and the correct handling of skewed continuous variables. I commend your excellent work and your receptiveness to the feedback.

I recommend "Minor Revisions" to address a few remaining formatting and consistency details that appear to have been overlooked in the final draft:

Section 3.1 (Text inconsistency): While Table 1 was correctly updated to show Medians [IQR], the narrative text in Section 3.1 still reports means for Age. Please update the text to align with the revised Table 1.

Table 1 (Typo): There is a minor typo in the newly added row; please change "PRISMA IV Score" to the correct term, "PRISM IV Score".

Figures 2 & 3 (Formatting): As suggested in the previous round, please move the floating exact p-values to the figure legends or replace them with standard asterisk notations to improve the visual clarity of the charts.

Once these minor corrections are made, the manuscript will be a valuable addition to the literature. I wish you the best of luck with this publication and your future research endeavors.

Author Response

Dear Reviewer,

Thank you very much for your careful re-evaluation of our revised manuscript and for your positive and encouraging comments. We are grateful for your recognition of the improvements made to the statistical analysis and reporting.

We have carefully addressed all remaining minor revisions as suggested:

Section 3.1: The narrative text has been revised to align with Table 1 by reporting age as median [IQR] rather than mean values.
Table 1: The typographical error has been corrected from “PRISMA IV Score” to “PRISM IV Score.”
Figures 2 and 3: The floating exact p-values have been removed from the figures and replaced with standard asterisk notation. The corresponding explanations have been included in the figure legends to improve clarity and readability.

We sincerely appreciate your constructive feedback, which has helped us further improve the consistency and presentation of the manuscript. Thank you again for your valuable time and thoughtful review.

Reviewer 4 Report

Comments and Suggestions for Authors

Dear authors, thank you for the new version. Most of my comments have been addressed. The AKI incidence has been corrected, the study period is clearer, the statistical results are presented more carefully, and the complications variable is better explained. But, One minor point remains. You note that data are unavailable regarding whether KDIGO staging was based on creatinine, urine output, or both, or whether the baseline creatinine was actual or estimated. Please state more clearly in the limitations section that this may affect the accuracy of the staging and the comparability of the results with other studies.

So, I suggest accept after minor revision.

Thank you

Comments on the Quality of English Language

The English could be improved to more clearly express the research.

Author Response

Dear Reviewer,
We sincerely thank you for your careful re-evaluation of the manuscript and for acknowledging that most of the previous concerns have been addressed. We appreciate your remaining suggestion regarding the limitations section.
We have revised the limitations section to state more explicitly that the dataset did not contain separate variables indicating whether final KDIGO staging was based on creatinine criteria, urine output criteria, or both, nor whether baseline creatinine was measured from prior records or estimated using the Schwartz formula. We now clarify that these limitations may affect the accuracy of AKI staging and may reduce direct comparability with studies that report criterion-specific staging or measured versus estimated baseline creatinine.
We have also improved the scientific English throughout the manuscript to express the study methods, findings, and limitations more clearly. Thank you again for your helpful comments, which have improved the clarity and balance of the manuscript.