Identifying a Safety Threshold for Parenteral Glucose Intake in the Early Acute Phase of Preterm Neonates

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This is a well-written and methodologically sound retrospective cohort study that addresses a clinically significant question in neonatology. In my view, the study is logically structured, the limitations are honestly acknowledged, and the discussion is appropriately cautious. The manuscript is suitable for publication after addressing the following points.

1, The central weakness, which the authors correctly identify, is that the 7 g/kg/day threshold is derived from and then tested on the same dataset. This approach is highly susceptible to overfitting and optimism bias, meaning the true risk inflection point may be different and the strength of the association is likely overestimated. While a formal external cohort is beyond the scope of a revision, the evidence of this association could be substantially strengthened by an internal validation technique. I request that the authors include a sensitivity analysis using k-fold cross-validation or, at a minimum, bootstrap resampling to internally validate the inflection point and the confidence intervals of the adjusted odds ratios in Figure 3. This would provide the reader with a more realistic assessment of the model's stability and predictive performance within the available data. The text in the limitations section should state explicitly that the results remain subject to optimism bias until externally validated.

2, The authors define their exposure as the total daily parenteral glucose intake in g/kg/day, arguing this is more clinically meaningful than an instantaneous infusion rate (GIR). This is a valid perspective, but it conflates the absolute glucose load with the rate of glucose oxidation. A neonate receiving 7.2 g/kg/day over 24 hours via continuous infusion is metabolically different from one receiving the same total daily dose but with periods of higher and lower rates. Were there any protocol violations, instances of temporarily stopped infusions that were then "caught up," or medications administered in dextrose-containing fluids (e.g., inotropes) that could have created unmeasured peaks in the glucose infusion rate, even within a single day's total? Please discuss this in the methods or limitations. A metric like the "area under the curve" for intake above a certain GIR, or the peak daily GIR, might better capture the mechanism of exceeding oxidative capacity. Please justify the chosen metric (total g/kg/day) more explicitly over the GIR, or, if data are available, consider presenting a secondary analysis using a time-averaged or peak GIR metric.

3, The model in Figure 3 simultaneously adjusts for 'ELBW' (<1000g) and 'GA <32 weeks'. These variables are collinear and both serve as markers of prematurity and vulnerability. Adjusting for both can lead to inflated standard errors and unstable effect estimates. A more informative model might use either GA (as a continuous or categorical variable) and BW z-score to represent the growth and maturity components separately. The authors should: (a) examine and report the variance inflation factor (VIF) between these two variables; (b) if collinearity is present, justify adjusting for both, or present an alternative model that separates these into non-collinear components of maturity and fetal growth.

4, line 29 to 31, "A safety threshold for parenteral glucose may exist…", is too vague. Please revise the abstract's conclusion to quantitatively summarize the key finding from the multivariable analysis (e.g., "In this single-center cohort, exceeding 7 g/kg/day of parenteral dextrose was independently associated with a [xx]% increase in the odds of hyperglycemia…, with no benefit on EUGR.") while maintaining the call for prospective validation.

5, line 118 to 120, the definition of hyperglycemia (>180 mg/dL on two consecutive measurements at least 3 hours apart) is pragmatic but could be influenced by measurement frequency. Were blood glucose measurements performed according to a uniform schedule, or was testing driven by clinical signs? A difference in surveillance frequency between the groups could introduce detection bias. Please clarify the glucose monitoring protocol in the NICU.

6, line 116 to 125, many of the paragraphs only have one sentence. Please combine if possible. 

Author Response

Comment 1. Internal validation and optimism bias

Reviewer 1, Comment 1. The central weakness, which the authors correctly identify, is that the 7 g/kg/day threshold is derived from and then tested on the same dataset. This approach is highly susceptible to overfitting and optimism bias, meaning the true risk inflection point may be different and the strength of the association is likely overestimated. While a formal external cohort is beyond the scope of a revision, the evidence of this association could be substantially strengthened by an internal validation technique. I request that the authors include a sensitivity analysis using k fold cross validation or, at a minimum, bootstrap resampling to internally validate the inflection point and the confidence intervals of the adjusted odds ratios in Figure 3. This would provide the reader with a more realistic assessment of the model stability and predictive performance within the available data. The text in the limitations section should state explicitly that the results remain subject to optimism bias until externally validated.

Authors' response.

We thank the Reviewer for this suggestion. As requested, we have performed an internal validation of the two multivariable models presented in Figure 3 using non parametric bootstrap resampling for bias corrected and accelerated (BCa) 95% confidence intervals on the adjusted odds ratios, and stratified 10 fold cross validation repeated 100 times for the apparent and cross validated Harrell C statistic, the corresponding optimism, and the cross validated calibration slope. Methodological details are reported in the revised Methods (§2.6) and the full numerical results in Supplementary Tables S2 and S3.

The internal validation supports the qualitative conclusions of the primary analyses: the BCa 95% confidence intervals for the adjusted odds ratio of exceeding the 7 g/kg/day threshold excluded the null in both models, the optimism in discrimination was modest, and the cross validated calibration slopes were close to unity. Per covariate estimates and discrimination/calibration metrics are reported in Supplementary Tables S2 and S3.

 

In agreement with the Reviewer, the Limitations section has been expanded to state explicitly that, even after internal validation, the magnitude of the associations may not generalise beyond the present cohort and that external validation in independent preterm populations remains required.

Comment 2. Total daily intake versus glucose infusion rate

Reviewer 1, Comment 2. The authors define their exposure as the total daily parenteral glucose intake in g/kg/day, arguing this is more clinically meaningful than an instantaneous infusion rate (GIR). This is a valid perspective, but it conflates the absolute glucose load with the rate of glucose oxidation. A neonate receiving 7.2 g/kg/day over 24 hours via continuous infusion is metabolically different from one receiving the same total daily dose but with periods of higher and lower rates. Were there any protocol violations, instances of temporarily stopped infusions that were then "caught up", or medications administered in dextrose containing fluids (e.g., inotropes) that could have created unmeasured peaks in the glucose infusion rate, even within a single day total? Please discuss this in the methods or limitations. A metric like the area under the curve for intake above a certain GIR, or the peak daily GIR, might better capture the mechanism of exceeding oxidative capacity. Please justify the chosen metric (total g/kg/day) more explicitly over the GIR, or, if data are available, consider presenting a secondary analysis using a time averaged or peak GIR metric.

Authors' response.

We thank the Reviewer for this thoughtful methodological observation, which has prompted us to clarify our choice of exposure metric. The total daily parenteral dextrose intake in g/kg/day was chosen for three converging reasons. First, international guidelines for the early acute phase express both the lower and the upper recommended boundaries in g/kg/day (Moltu et al., 2021), so the exposure metric is directly aligned with the metric on which the prescribing clinician acts. Second, the pathophysiological mechanism we are testing, namely saturation of the limited glucose oxidative capacity in the early acute phase, is a function of cumulative load over 24 hours rather than of an instantaneous rate. Third, the studies that have reported a positive association between parenteral glucose and hyperglycaemia (Stensvold 2015, Zamir 2019, Tottman 2018) all used cumulative intake, whereas the study that did not find an association (Beardsall NIRTURE 2010) used a time point GIR; we have strengthened this point in §4.2.

We also wish to reassure the Reviewer regarding unmeasured intraday peaks. In our NICU, parenteral nutrition was compounded daily and prescribed in g/kg/day; catch up infusions were not allowed, and whenever an interruption of parenteral nutrition longer than 10 minutes was required to administer a medication, a 10% dextrose infusion was simultaneously started through a separate venous access at the same glucose rate as the ongoing parenteral nutrition, so that the effective rate of glucose delivery was kept stable. Vasoactive drugs and other parenteral medications were diluted in 5% dextrose at low volumes, contributing a negligible additional glucose load. These details, previously implicit in routine clinical practice, are now made explicit in Methods §2.4. As a consequence, the effective intraday GIR was substantially stable by protocol, and the total daily intake provides a faithful representation of the cumulative daily glucose load.

We acknowledge, as a residual limitation, that minor unmeasured fluctuations in instantaneous GIR cannot be entirely excluded in a retrospective design, and we have stated this explicitly in §4.8.

Comment 3. Collinearity between ELBW and gestational age below 32 weeks

Reviewer 1, Comment 3. The model in Figure 3 simultaneously adjusts for ELBW (<1000 g) and GA <32 weeks. These variables are collinear and both serve as markers of prematurity and vulnerability. Adjusting for both can lead to inflated standard errors and unstable effect estimates. A more informative model might use either GA (as a continuous or categorical variable) and BW z score to represent the growth and maturity components separately. The authors should: (a) examine and report the variance inflation factor (VIF) between these two variables; (b) if collinearity is present, justify adjusting for both, or present an alternative model that separates these into non collinear components of maturity and fetal growth.

Authors' response.

We thank the Reviewer for this important point. ELBW (<1000 g) and gestational age <32 weeks were retained as covariates because they capture two distinct dimensions of preterm vulnerability, fetal growth and maturity respectively; in our cohort the correlation between the two binary descriptors was modest (phi 0.225 in the hyperglycaemia model, 0.232 in the hypertriglyceridaemia model). Formal collinearity diagnostics supported their simultaneous inclusion: variance inflation factors were ≤1.40 for both variables in both models, well below the conventional threshold of 5, and the condition number of the design matrix was 6.25 and 6.13 respectively, well below the threshold of 30. This diagnostic is now made explicit in Methods section.

Comment 4. Quantitative summary of the principal finding in the Abstract

Reviewer 1, Comment 4. Line 29 to 31, "A safety threshold for parenteral glucose may exist…", is too vague. Please revise the abstract conclusion to quantitatively summarize the key finding from the multivariable analysis (e.g., "In this single center cohort, exceeding 7 g/kg/day of parenteral dextrose was independently associated with a [xx]% increase in the odds of hyperglycemia…, with no benefit on EUGR.") while maintaining the call for prospective validation.

Authors' response.

The closing sentences of the Abstract have been revised to provide a quantitative summary of the principal multivariable findings, while preserving the call for prospective validation, as suggested.

Comment 5. Glucose monitoring schedule and detection bias

Reviewer 1, Comment 5. Line 118 to 120, the definition of hyperglycemia (>180 mg/dL on two consecutive measurements at least 3 hours apart) is pragmatic but could be influenced by measurement frequency. Were blood glucose measurements performed according to a uniform schedule, or was testing driven by clinical signs? A difference in surveillance frequency between the groups could introduce detection bias. Please clarify the glucose monitoring protocol in the NICU.

Authors' response.

We thank the Reviewer for raising this point. In our NICU, blood glucose monitoring during the first week of life follows a protocol-driven schedule. During the first 72 hours of life, blood glucose was measured every 6 hours in all neonates receiving parenteral nutrition; whenever a value exceeded 180 mg/dL, the measurement was repeated 3 hours later, and hyperglycaemia was diagnosed only if the second measurement also exceeded 180 mg/dL. After the first 72 hours, blood glucose was measured every 12 hours. Because this surveillance schedule was uniform across the cohort and predefined by day of life rather than by exposure or by clinical condition, differential surveillance frequency between exposed and unexposed neonates is unlikely to have driven the observed associations. This protocol has been described in greater detail in the revised Methods.

Comment 6. Editorial: paragraph consolidation

Reviewer 1, Comment 6. Line 116 to 125, many of the paragraphs only have one sentence. Please combine if possible.

Authors' response.

We have consolidated the relevant section. Specifically, the four single sentence paragraphs containing the operational definitions of hyperglycaemia, hypertriglyceridaemia, metabolic acidosis and EUGR have been merged into one cohesive paragraph defining the primary outcomes, followed by a separate paragraph for the secondary outcome. The resulting text retains all the original content with no loss of information and improves readability as suggested.

 

Reviewer 2 Report

Comments and Suggestions for Authors

Dear Authors, 

I highly appreciate your interest and effort to identify optimization ways to address to this critical issue in newbornes management.

I only few minor suggestions for the best interest of scientific rigor and future research and clinical perspectives of the study.

1. The abstract clearly summarizes the findings. The introduction provides a strong rationale, highlighting the metabolic vulnerability of preterm neonates in the "early acute phase". In this section, please clarify the "late acute" and "recovery" phases briefly to contextualize why the first week (early acute) is uniquely dangerous for glucose over-provision.
2. The exclusion of neonates who died or were transferred within 72 hours is appropriate to ensure sufficient PN exposure data, but may introduce a survivor bias that should be briefly discussed.
3. I observed that you are cautious in your conclusions, insisting that prospective multicenter studies are required, which I concur with.
4.  You  correctly acknowledge that the 7~g/kg/day threshold was identified "in-sample," which may lead to optimism bias. Please provide a brief recommendation on how clinicians should monitor infants once they approach the 7~g/kg/day mark.
5. Table 1 provides a comprehensive overview of the population. However, you should confirm if the use of "actual intake" rather than "prescribed" led to significant missing data in cases where nursing charts were incomplete
6. Validation: If possible, suggest how this threshold might be validated in a separate subset of the data or a different cohort.

Author Response

We thank the Reviewer for the favourable evaluation and for the constructive minor suggestions. We respond to each point individually below.

Comment 1. Brief definition of late acute and recovery phases

Reviewer 2, Comment 1. The abstract clearly summarizes the findings. The introduction provides a strong rationale, highlighting the metabolic vulnerability of preterm neonates in the early acute phase. In this section, please clarify the late acute and recovery phases briefly to contextualize why the first week (early acute) is uniquely dangerous for glucose over provision.

Authors' response.

We thank the Reviewer for this thoughtful suggestion. The Introduction has been revised as suggested, with a brief description of the late acute and recovery phases added (§1) to contextualise why the early acute phase is uniquely vulnerable to glucose over-provision.

Comment 2. Survivor bias from the 72 hour exclusion

Reviewer 2, Comment 2. The exclusion of neonates who died or were transferred within 72 hours is appropriate to ensure sufficient PN exposure data, but may introduce a survivor bias that should be briefly discussed.

Authors' response.

We thank the Reviewer for this fair observation. The 72-hour eligibility criterion was necessary to ensure that the exposure window of interest could be observed, but we agree that it may introduce a survivor bias. A brief acknowledgement of this point has been added to the Limitations section, noting that the cohort is representative of preterm neonates who survive at least the first three days of life and who are not transferred elsewhere in the same window.

Comment 3. Caution of the conclusions

Reviewer 2, Comment 3. I observed that you are cautious in your conclusions, insisting that prospective multicenter studies are required, which I concur with.

Authors' response.

We thank the Reviewer for this supportive observation. The cautionary tone has been preserved throughout the revised manuscript.

Comment 4. Practical recommendation for monitoring near 7 g/kg/day

Reviewer 2, Comment 4. You correctly acknowledge that the 7 g/kg/day threshold was identified in sample, which may lead to optimism bias. Please provide a brief recommendation on how clinicians should monitor infants once they approach the 7 g/kg/day mark.

Authors' response.

We thank the Reviewer for this thoughtful question. Our data support a prudent approach: until more robust evidence becomes available, parenteral dextrose intake during the early acute phase should not exceed the 7 g/kg/day threshold identified in our cohort. Beyond the first week of life, current guidelines do not specify how parenteral glucose intake should be subsequently advanced; in a previous randomised controlled trial from our group, a delayed achievement of macronutrient targets was associated with a lower incidence of hyperglycaemia in very low birth weight neonates (Di Chiara et al., Nutrients, 2023). After the first week, the increasing metabolic stability of the neonate and the concurrent progression of enteral nutrition may further mitigate the metabolic burden of parenteral glucose support.

Comment 5. Actual versus prescribed intake and missing data

Reviewer 2, Comment 5. Table 1 provides a comprehensive overview of the population. However, you should confirm if the use of "actual intake" rather than "prescribed" led to significant missing data in cases where nursing charts were incomplete.

Authors' response.

We thank the Reviewer for this thoughtful question. In our unit the volume of parenteral nutrition actually administered was recorded contemporaneously on a structured chart and cross-checked daily by the attending neonatologist against the prescription and the residual volume in the infusion bag. As a consequence, the actual daily parenteral dextrose intake was retrievable for every neonate on every day of the first week of life, and the use of "actual intake" rather than "prescribed" did not introduce missing data on the exposure of interest.

Comment 6. Validation of the threshold in a separate subset or different cohort

Reviewer 2, Comment 6. Validation: If possible, suggest how this threshold might be validated in a separate subset of the data or a different cohort.

Authors' response.

We thank the Reviewer for this important suggestion. We address it on three complementary levels.

(1) The ideal design for external validation. External validation of the 7 g/kg/day threshold should be carried out in a multicentre prospective cohort of preterm neonates with gestational age at or below 34 weeks or birth weight at or below 1500 grams, recruited under harmonised parenteral nutrition protocols and with prospectively recorded daily parenteral and enteral macronutrient intakes during the first week of life. The same operational definitions of hyperglycaemia, hypertriglyceridaemia and EUGR adopted in the present study should be applied. The threshold could be evaluated as a fixed cut-off (intake exceeding 7 g/kg/day on at least one day of the first week of life) and, in parallel, re-estimated by spline regression in the validation cohort, to assess both transportability and stability. A nested individual patient data meta-analytic framework, pooling existing prospective preterm cohorts, would represent the most efficient route to confirmation.

(2) Acknowledgement of the limitation. The need for external validation is explicitly stated in the Limitations section (§4.8): the strength of the association between exceeding the 7 g/kg/day threshold and the metabolic outcomes may not generalise beyond the present cohort, and external validation in independent preterm populations is therefore required before the threshold can be applied as a definitive risk-stratification cut-off.

(3) Mitigation within the available data. To partially mitigate this limitation, we have implemented an internal validation strategy (as detailed in our response to Reviewer 1, Comment 1) based on stratified non-parametric bootstrap resampling (1,000 resamples) for bias-corrected and accelerated 95% confidence intervals of each adjusted odds ratio, and on stratified 10-fold cross-validation repeated 100 times for the apparent and cross-validated Harrell C statistic and calibration slope of each multivariable model. Methodological details are reported in Methods §2.6 and full numerical results in Supplementary Tables S2 and S3.

Reviewer 3 Report

Comments and Suggestions for Authors

The first days after preterm birth are critical.  Besides ensuring gas exchange, the other major factor neonatologists can address is nutrition support.  Despite nearly 75 years of providing nutrition support to infants surviving birth at 32 weeks and less, there remain numerous gaps in knowledge surrounding the nutritional needs of extremely preterm infants.  Surprisingly, this includes understanding the provision of glucose during the first week after birth.  From this perspective, the authors’ contribution is valuable and will contribute to improving the care of newborn preterm infants.  There are some points the authors may wish to consider to improve their contribution.  Many, maybe most, of the following are about minor points.

 

1. The term “control” infers all were provided the same parenteral glucose intake. That isn’t the case.  Perhaps terms that indicate “high” and “low” parenteral glucose would be more appropriate. 
2. Line 44: Based on the exclusion criteria (lines 92 -96), critically ill preterm infants were not included.  The study population was composed of extremely preterm and SGA infants who are at “critical” and at risk but not necessarily “ill”.   It is not clear if infants with RDS, NEC, sepsis or other post-delivery complications were excluded.  If not, were those conditions considered in the analysis?  This is important because of the impacts on glucose metabolism, notably hyperglycemia (Butorac Ahel I, et al. Incidence and Risk Factors for Glucose Disturbances in Premature Infants. Medicina (Kaunas). 2022 Sep 16;58(9):1295. doi: 10.3390/medicina58091295). If such infants were excluded, this should be explained. 
3. Curious” Do the authors know or have an estimate of the daily provision of glucose from the placenta to the fetus?  
4. Ls 99-104: Was the glucose provided by enteral nutrition added to that provided parenterally?  Based on lines 143-148 enteral glucose was considered.  If so, this should be emphasized.  Also, the amount of glucose in breast is less than what is available in preterm formulas because breast milk has lactose whereas preterm formulas have a combination of lactose and glucose polymers that have a higher glycemic index.  This should be considered to avoid potential criticisms.
5. Can the authors provide more details about the PN protocol? Was only glucose provided?  Or did PN include amino acids and lipids?  The focus is on glucose, but some readers will want to know if PN was complete. 
6. The authors should include information about the volumes of PN and EN provided during the 7-day study period. Did they differ between the low and high groups?  Was this considered in the stat analysis?  This is important because of splanchnic metabolism of glucose (van der Schoor SR. Splanchnic bed metabolism of glucose in preterm neonates. Am J Clin Nutr. 2004 May;79(5):831-7).
7. What about the total amounts of glucose provided parenterally during the study period?
8. L 179: the term imputed is not appropriate.
9. Do the authors have any information about insulin delivery in response to hyperglycemia?
10. If peak parenteral glucose in excess of 7 g/kg-d was transient, was the associated hyperglycemia and hyperlipidemia also transient?
11. Ls 563-566: why wasn’t it possible to correlate when parenteral glucose exceeded 7 g/kg-d and onset of hyperglycemia?
12. Section 4.8 could include thoughts about future efforts.

 

Additional thoughts:

• Would future studies benefit from the use of continuous glucose monitors?
• Because of the limitations associated with preterm infants, would a large animal model allow the authors to address in a very controlled manner specific aspects of the relationship between parenteral glucose and hyperglycemia/hyperlipidemia?

Author Response

We thank the Reviewer for the favourable overall assessment of our contribution and for the thoughtful and constructive comments. We respond to each point individually below; comments 1 to 12 follow the numbering used by the Reviewer, and the two additional thoughts are addressed at the end.

Comment 1. Use of the term "control"

Reviewer 3, Comment 1. The term "control" infers all were provided the same parenteral glucose intake. That isn’t the case. Perhaps terms that indicate "high" and "low" parenteral glucose would be more appropriate.

Authors' response.

We thank the Reviewer for raising this point. In our manuscript, the terms "cases" and "controls" are used in the conventional epidemiological sense for observational analyses: cases are neonates who developed the metabolic complication of interest, and controls are neonates who did not. The two groups are therefore defined by the outcome and not by the exposure. The legend of Figure 2 already makes this operational definition explicit; to prevent any further ambiguity, the corresponding sentence in Results §3.3.2 has been expanded at first mention to recall the same definition.

Comment 2. Critically ill infants and post-delivery complications

Reviewer 3, Comment 2. Line 44: Based on the exclusion criteria (lines 92–96), critically ill preterm infants were not included. The study population was composed of extremely preterm and SGA infants who are at "critical" and at risk but not necessarily "ill". It is not clear if infants with RDS, NEC, sepsis or other post-delivery complications were excluded. If not, were those conditions considered in the analysis? This is important because of the impacts on glucose metabolism, notably hyperglycemia (Butorac Ahel I, et al. Incidence and Risk Factors for Glucose Disturbances in Premature Infants. Medicina (Kaunas). 2022 Sep 16;58(9):1295. doi: 10.3390/medicina58091295). If such infants were excluded, this should be explained.

Authors' response.

We thank the Reviewer for this important point and for the relevant reference (Butorac Ahel et al., Medicina, 2022). To clarify, in our cohort post-delivery complications such as respiratory distress syndrome, necrotising enterocolitis, culture-proven sepsis, intraventricular haemorrhage, bronchopulmonary dysplasia and retinopathy of prematurity were not exclusion criteria, and neonates affected by these conditions were included in the analytic cohort. The composite covariate "neonatal morbidity" adjusted for in the multivariable models (Figure 3) was specifically defined to capture the morbidities with the highest a priori relevance for glucose metabolism within the exposure window of interest (postnatal days 0 to 7), namely culture-proven sepsis (acting through systemic inflammation and insulin resistance) and intraventricular haemorrhage (acting through haemodynamic and stress responses). Bronchopulmonary dysplasia and retinopathy of prematurity were not included, since they are diagnosed at 36 weeks postmenstrual age and after several weeks of postnatal life respectively, and therefore cannot represent confounders of the exposure-outcome association observed in the first week of life. Necrotising enterocolitis has a typical presentation beyond the first week of life. Respiratory distress syndrome, virtually universal in very preterm neonates, was not modelled as a separate covariate because it is collinear with prematurity, which is already captured by gestational age <32 weeks and ELBW. The resulting covariate set is consistent with the principal predictors of glucose disturbances identified by Butorac Ahel et al. (gestational age, birth weight and sepsis). The operational definition of the "neonatal morbidity" composite covariate is now made explicit in Methods §2.6 of the revised manuscript.

Comment 3. Daily placental glucose provision to the fetus

Reviewer 3, Comment 3. Curious: do the authors know or have an estimate of the daily provision of glucose from the placenta to the fetus?

Authors' response.

We thank the Reviewer for this thoughtful question. Published estimates indicate that the human fetus in the third trimester receives glucose from the placenta at approximately 5 to 8 mg/kg/min (Hay, Neonatology, 2006; Sunehag et al., Pediatr Res, 1999), corresponding to about 7 to 12 g/kg/day. Historically, the recommendations for parenteral glucose intake in the preterm neonate were anchored to this fetal-equivalent provision, on the assumption that the ex-utero neonate could be treated as a continuation of the in-utero fetus; this is the rationale that supported the so-called "aggressive" parenteral nutrition protocols, which advocated glucose infusion rates up to 12.5 mg/kg/min (approximately 18 g/kg/day) from the first hours of life. Evidence accumulated over the past decade has progressively challenged this paradigm. Our group, in a randomised controlled trial, demonstrated that delayed achievement of macronutrient targets substantially reduces the incidence of hyperglycaemia and is associated with better growth at 12 months of corrected age in very low birth weight neonates (Di Chiara et al., Nutrients, 2023). The same group has also reported that parenteral-nutrition-related hyperglycaemia adversely affects long-term neurodevelopment in preterm newborns (Boscarino et al., Nutrients). Independent prospective cohort and before-after studies have converged on similar findings (Stensvold et al., JAMA Pediatr, 2015; Stensvold et al., Acta Paediatr, 2018; Zamir et al., J Pediatr, 2019; Tottman et al., JPEN, 2018). Taken together, these data indicate that the preterm neonate ex utero is not metabolically equivalent to a fetus of the same gestational age: hepatic endogenous glucose production is not completely suppressed by exogenous infusion, peripheral insulin sensitivity is reduced, and the pancreatic insulin response is immature. The 5 to 8 g/kg/day range proposed by ESPGHAN/ESPEN/ESPR/CSPEN for the early acute phase (Moltu et al., 2021) already reflects this downward revision relative to the fetal-equivalent target. Our finding that the inflection point lies within this range, at 7 g/kg/day, is consistent with the position that, during the early acute phase, the preterm neonate tolerates less parenteral glucose than the fetus would receive in utero.

Comment 4. Enteral glucose and glycemic load of breast milk versus preterm formula

Reviewer 3, Comment 4. Lines 99–104: was the glucose provided by enteral nutrition added to that provided parenterally? Based on lines 143–148 enteral glucose was considered. If so, this should be emphasized. Also, the amount of glucose in breast milk is less than what is available in preterm formulas because breast milk has lactose whereas preterm formulas have a combination of lactose and glucose polymers that have a higher glycemic index. This should be considered to avoid potential criticisms.

Authors' response.

We thank the Reviewer for raising this point. The exposure of interest in our study is specifically the parenteral dextrose intake (§2.2), not the total daily glucose intake. The rationale is threefold: physiologically, parenteral glucose bypasses the first-pass hepatic-intestinal modulation and enters the systemic circulation directly, with a metabolic profile distinct from that of enteral glucose; clinically, the ESPGHAN/ESPEN/ESPR/CSPEN recommendations for the early acute phase are expressed in g/kg/day of parenteral glucose, not total glucose, so the exposure is aligned with the prescribing metric; and pragmatically, parenteral glucose is the component over which the clinician has direct prescriptive control during the early postnatal period. Enteral nutrition is nonetheless considered in two ways. First, the local protocol (§2.4) prescribes minimal enteral feeding initiated within 48 h after birth at 10-20 mL/kg/day and progressively advanced by 20-30 mL/kg/day as tolerated, with fresh unfortified human milk from the infant’s own mother as the first choice and preterm formula reserved for cases where human milk is unavailable or insufficient. Second, the multivariable models (Figure 3) include "delayed enteral nutrition" (enteral feeding initiated after 72 h of life) as a covariate, which captures the residual confounding related to the enteral route. Quantitatively, in our cohort the mean enteral volume over the first week of life was approximately 19 to 34 mL/kg/day (median 19.4 mL/kg/day); assuming a carbohydrate content of approximately 7 g per 100 mL (both for human milk lactose and for preterm formula), this corresponds to a mean enteral carbohydrate load of approximately 0.9 to 1.6 g/kg/day, that is, about 10 to 20% of the peak parenteral dextrose load observed in the cohort. The contribution of enteral glucose to the overall carbohydrate load is therefore quantitatively limited during the exposure window of the study. We also acknowledge the Reviewer’s observation regarding the higher glycaemic load of preterm formulas (combination of lactose and glucose polymers) compared with mature human milk (lactose only). In our cohort, fresh unfortified human milk from the infant’s own mother was offered as the first choice; however, as is typical in preterm birth, the availability of mother’s own milk in the first days of life is variable depending on maternal lactation, and preterm formula was used as a supplement when needed.

Comment 5. Composition of the parenteral nutrition protocol

Reviewer 3, Comment 5. Can the authors provide more details about the PN protocol? Was only glucose provided? Or did PN include amino acids and lipids? The focus is on glucose, but some readers will want to know if PN was complete.

Authors' response.

We thank the Reviewer for this important point. Parenteral nutrition in our NICU was complete and included, in addition to glucose (dextrose), parenteral amino acids and intravenous lipid emulsion, together with electrolytes (sodium, potassium, calcium, magnesium, phosphorus), formulated daily by the attending neonatologist based on each infant’s clinical condition, body weight and laboratory findings, and compounded by the hospital pharmacy. According to the local protocol, parenteral amino acids were typically initiated at approximately 1.0 g/kg/day on the first day of life and progressively advanced to approximately 2.5 g/kg/day by the end of the first week; intravenous lipid emulsion was typically initiated at approximately 0.3 g/kg/day and progressively advanced to approximately 1.8 g/kg/day. This information has now been added to Methods §2.4. We selected parenteral glucose as the exposure of interest because of its specific pathophysiological rationale during the early acute phase (saturation of the oxidative capacity and incomplete suppressibility of endogenous hepatic glucose production) and because the ESPGHAN/ESPEN/ESPR/CSPEN recommendations for this phase are expressed in g/kg/day of parenteral glucose.

Comment 6. Volumes of PN and EN during the 7-day study period

Reviewer 3, Comment 6. The authors should include information about the volumes of PN and EN provided during the 7-day study period. Did they differ between the low and high groups? Was this considered in the statistical analysis? This is important because of splanchnic metabolism of glucose (van der Schoor SR. Splanchnic bed metabolism of glucose in preterm neonates. Am J Clin Nutr. 2004 May;79(5):831–837).

Authors' response.

We thank the Reviewer for this important point and for the relevant reference (van der Schoor SR, Am J Clin Nutr, 2004). Total fluid intake (parenteral plus enteral) was pre-specified by the unit protocol and progressively advanced from 70 to 180 mL/kg/day across the first week of life, irrespective of the exposure group (§2.4); in our cohort the cumulative total volume received over the first week was therefore similar between neonates exceeding 7 g/kg/day of parenteral dextrose and neonates remaining at or below the threshold. In line with the physiological rationale highlighted by van der Schoor et al., who demonstrated that splanchnic-bed metabolism modulates the systemic impact of enteral glucose, we considered enteral nutrition in the analysis: the multivariable models of Figure 3 include "delayed enteral nutrition" (enteral feeding initiated after 72 h of life) as a covariate, capturing the contribution of the enteral route to the exposure-outcome association.

Comment 7. Total parenteral glucose during the study period

Reviewer 3, Comment 7. What about the total amounts of glucose provided parenterally during the study period?

Authors' response.

We thank the Reviewer for this point. In our cohort the total cumulative parenteral glucose received during the first week of life was, on average, 48.08 ± 28.77 g/kg (median 43.90, interquartile range 25.34 to 63.52, range 1.52 to 147.14 g/kg). This information has been added to the Results section of the revised manuscript as suggested.

Comment 8. Use of the term "imputed"

Reviewer 3, Comment 8. Line 179: the term "imputed" is not appropriate.

Authors' response.

We thank the Reviewer for this terminological observation. The sentence in Methods §2.6 has been rephrased for clarity to avoid the technical term "imputed" and to convey the methodological information in a more transparent way. The revised sentence now reads: "No imputation procedure was applied to missing data; complete-case analysis was performed for each multivariable model."

Comment 9. Insulin delivery in response to hyperglycemia

Reviewer 3, Comment 9. Do the authors have any information about insulin delivery in response to hyperglycemia?

Authors' response.

We thank the Reviewer for this point. The management of confirmed hyperglycaemia in our NICU followed the ESPGHAN/ESPEN/ESPR/CSPEN recommendations for parenteral nutrition in preterm neonates (Moltu et al., 2021), with a stepwise approach starting from a progressive reduction of the parenteral glucose intake and reserving insulin infusion as a second-line strategy when blood glucose remained above 180 mg/dL despite glucose reduction. In our cohort, 38 of 371 neonates (10.2%) received insulin during the first week of life. This information has been added to Methods §2.4.

Comment 10. Transient excess of glucose and transient metabolic complications

Reviewer 3, Comment 10. If peak parenteral glucose in excess of 7 g/kg/day was transient, was the associated hyperglycemia and hypertriglyceridemia also transient?

Authors' response.

The primary metabolic outcomes of our study were defined as the occurrence of hyperglycaemia or hypertriglyceridaemia during the first week of life, that is, the early acute phase that represents the most critical metabolic window for the preterm neonate. By their nature, these complications during the early acute phase are typically transient. As now better specified in the Limitations section (§4.8), the temporal overlap between the exposure and the outcome within the first week of life precludes a precise individual-level assessment of the correspondence between the duration of the exceedance and the duration of the complication.

Comment 11. Temporal correlation between exceeding 7 g/kg/day and onset of hyperglycemia

Reviewer 3, Comment 11. Lines 563–566: why was it not possible to correlate the day when parenteral glucose exceeded 7 g/kg/day and the onset of hyperglycemia?

Authors' response.

We thank the Reviewer for this important point. To partially address the methodological limitation arising from the overlap between the exposure and the outcome windows in the first week of life, we performed a day-by-day analysis of the parenteral glucose intake stratified by the occurrence of each metabolic complication (Supplementary Figure S1). This analysis showed that the parenteral glucose intake of cases was already higher than that of controls from postnatal day 2 onwards, providing aggregate-level evidence that the increase of parenteral glucose intake preceded or coincided with the onset of the metabolic complication. A precise individual-level temporal correlation between the day of exceedance and the day of first hyperglycaemic episode is, however, precluded by the heterogeneity of the temporal resolution of the data (daily aggregation of the parenteral intake versus 6-hourly blood glucose measurements in the first 72 hours of life and 12-hourly thereafter) and would require a prospective time-to-event design, as acknowledged in the Limitations section.

Comment 12. Section 4.8: thoughts about future efforts

Reviewer 3, Comment 12. Section 4.8 could include thoughts about future efforts.

Authors' response.

A short paragraph on future research directions has been added at the end of §4.8 of the revised manuscript, outlining four lines of further investigation that we consider most relevant: prospective multicentre external validation of the 7 g/kg/day threshold; a prospective time-to-event design supported by continuous glucose monitoring; complementary preclinical studies in large animal models of prematurity; and long-term follow-up of preterm neonates exposed to different parenteral glucose intakes during the early acute phase, in line with our previous work on neurodevelopmental outcomes. This paragraph also addresses, in an integrated form, the two additional points raised by the Reviewer at the end of the comments (use of continuous glucose monitors in future studies and use of large animal models).

Additional point 1. Continuous glucose monitors in future studies

Reviewer 3, Additional point 1. Would future studies benefit from the use of continuous glucose monitors?

Authors' response.

We fully agree with the Reviewer. Continuous glucose monitoring would substantially improve the temporal resolution of glycaemic ascertainment and would enable a prospective time-to-event analysis of the relationship between parenteral glucose intake and hyperglycaemia at the individual level. The integration of continuous glucose monitoring in future prospective studies has been explicitly mentioned in the new paragraph on future research directions added at the end of §4.8.

Additional point 2. Large animal model

Reviewer 3, Additional point 2. Because of the limitations associated with preterm infants, would a large animal model allow the authors to address in a very controlled manner specific aspects of the relationship between parenteral glucose and hyperglycemia/hypertriglyceridemia?

Authors' response.

We fully agree with the Reviewer. Complementary preclinical studies in large animal models of prematurity would allow a controlled mechanistic dissection of specific aspects of the relationship between parenteral glucose load, oxidative capacity, and the development of hyperglycaemia and hypertriglyceridaemia, which is not feasible in the clinical setting. The role of large animal models has been explicitly mentioned in the new paragraph on future research directions added at the end of §4.8.

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors have revised the manuscript accordingly. It can be considered for publication.