Hypoxic-Ischaemic Encephalopathy in Maternal Diabetes: Mirror of a Chronic Antepartum Process or an Acute Intrapartum Event? A Systematic Review

Abstract

Background: Maternal diabetes is associated with several adverse perinatal outcomes, but its relationship with hypoxic-ischaemic perinatal injury remains poorly defined. Fetal compromise may reflect an acute intrapartum event, a chronic antepartum process, or both. Objective: To systematically review the evidence linking gestational diabetes mellitus (GDM) and pregestational diabetes to hypoxic-ischaemic perinatal injury, assessing diagnosed or coded neonatal brain injury, acid–base status at birth, and chronic antepartum markers separately. Methods: PubMed, Embase, and Scopus were searched from inception to 16 October 2026. Observational studies comparing pregnancies with and without diabetes and reporting diagnosed or coded hypoxic-ischaemic encephalopathy, neonatal encephalopathy or neonatal seizures, umbilical acid–base status, or a chronic marker of fetal hypoxia were eligible; studies reporting only Apgar-based or other non-specific outcomes were excluded. Risk of bias was assessed with the Newcastle–Ottawa Scale together with a separate judgement of outcome-definition adequacy, certainty of evidence was rated with GRADE, and synthesis followed the SWiM guideline, with vote counting by direction of effect. The protocol was registered in PROSPERO (CRD420261435362). Results: Seventy-four studies were included (30 GDM only, 12 pregestational diabetes only, 32 both or unspecified). Pregestational diabetes was associated with hypoxic-ischaemic encephalopathy, neonatal encephalopathy, or neonatal seizures in all five studies with sufficient events (adjusted odds or risk ratios 1.5–5.1; very low certainty). In GDM, estimates pointed in an adverse direction in six of seven informative comparisons but were mostly small and imprecise; statistically significant associations came only from a large registry (adjusted relative risk 1.24–1.25) and a study restricted to small-for-gestational-age infants (very low certainty). Umbilical acid–base status was less favourable in pregestational diabetes in 9 of 10 studies. In GDM, differences were small (less favourable in 13 of 17 comparisons; pooled mean difference in umbilical arterial pH −0.025, 95% CI −0.041 to −0.010, prediction interval −0.10 to 0.05, three studies; very low certainty for clinically relevant acidaemia). Chronic markers pointed in an adverse direction in 33 of 37 independent evidence sources, most consistently erythropoietin, nucleated red cells, and polycythaemia, but certainty was very low for every marker. No study documented a chronic marker, an identified acute event, and diagnosed or coded brain injury in the same fetus. Conclusions: Pregestational diabetes is associated with diagnosed or coded hypoxic-ischaemic encephalopathy, neonatal encephalopathy, or seizures and with markers of chronic fetal hypoxic stress, but the certainty of this evidence is very low. In GDM, estimates for these outcomes mostly point in an adverse direction but are small and imprecise (very low certainty), and acid–base differences at birth are small. Whether a chronic process and an acute event combine to cause injury in an individual fetus has not been studied.

1. Introduction

Gestational diabetes mellitus (GDM) affects a growing proportion of pregnancies worldwide, and its prevalence continues to rise alongside maternal age and obesity [1,2]. It is associated with fetal macrosomia, hypertensive disorders, and increased rates of operative delivery [1,2]. Its relationship with hypoxic-ischaemic perinatal injury, however, is still less clear, and this relationship carries substantial clinical and medico-legal weight [3,4]. Hypoxic-ischaemic encephalopathy (HIE) is often attributed to a single acute intrapartum event, since intrapartum monitoring naturally focuses attention there [5,6,7]. Yet fetal compromise may instead reflect a chronic, antepartum process. Chronic hyperglycaemia and hyperinsulinaemia can increase fetal oxygen demand, and they can also alter placental structure and function [1,2,8]. The proposed pathway can be summarised in sequence: maternal hyperglycaemia produces fetal hyperglycaemia and hyperinsulinaemia; fetal hyperinsulinaemia raises oxidative metabolism and oxygen consumption; in parallel, placental villous immaturity, a thicker diffusion barrier, and an altered feto-placental weight ratio may limit oxygen transfer; the resulting imbalance between oxygen supply and demand would be signalled by erythropoietin release, nucleated red cells, and polycythaemia; and the reduced fetal reserve would leave the fetus less able to tolerate the transient hypoxaemia of normal labour or an acute intrapartum event. A fetus exposed to this chronic burden may tolerate an acute peripartum event poorly, even when that event would be innocuous in an otherwise healthy pregnancy. Distinguishing an acute cause from a chronic, cumulative one is therefore central to any causal assessment of perinatal brain injury in diabetic pregnancy, and it is rarely straightforward in practice. Existing evidence on this question is fragmented. Studies vary widely in diabetes definition, in comparator selection, and in the hypoxia-related outcome they report. Some measure intrapartum cardiotocography or umbilical cord gases [2,9]. Others measure antepartum markers such as Doppler velocimetry, placental histology, or erythropoietin, but few measure more than one of these domains in the same cohort [10]. Moreover, much of the mechanistic evidence on chronic fetal hypoxia originates from pregestational diabetes rather than from GDM itself. Pregestational diabetes precedes conception, so it can demonstrate genuinely pre-existing vascular pathology, whereas GDM-related changes must, by definition, develop within the index pregnancy. The two conditions differ in at least four respects relevant to this question. First, exposure timing and duration: pregestational diabetes exposes the embryo and fetus from conception, whereas GDM is usually diagnosed in the late second or third trimester. Second, maternal vasculopathy: microvascular disease may pre-date pregnancy in type 1 and type 2 diabetes but is absent, by definition, in GDM. Third, metabolic exposure: hyperglycaemia is typically more severe and more variable in type 1 diabetes and includes periconceptional and first-trimester hyperglycaemia. Fourth, the meaning of a chronic hypoxic marker: in pregestational diabetes, it may reflect pre-existing placental vascular pathology, whereas in GDM, it can only reflect changes acquired within the index pregnancy. Evidence from pregestational diabetes therefore cannot be transferred to GDM without explicit justification, and this review reports the two populations separately throughout. To our knowledge, no previous systematic review has synthesised diagnosed or coded neonatal brain injury, acid–base status at birth, and chronic antepartum markers within a single, methodologically appraised review. In particular, it is unknown whether any study has documented each step of the proposed pathway (a chronic marker, an acute intrapartum event, and diagnosed or coded brain injury) in the same fetus. This gap limits the ability of clinicians and of those assessing causation after the fact to weigh the available evidence properly, and it leaves individual case assessments without a clear evidentiary framework. We therefore conducted a systematic review of the literature on maternal diabetes, both gestational and pregestational, and hypoxic-ischaemic perinatal injury. We assessed acute peripartum biochemical and clinical evidence, diagnosed or coded neurological outcome and follow-up, and chronic antepartum hypoxia markers, alongside the methodological quality of each included study. Our aim was to characterise, systematically and transparently, what the existing evidence does and does not support. To structure this assessment, we defined two working hypotheses. Under Hypothesis A, hypoxic-ischaemic injury results from an acute intrapartum event in a previously uncompromised fetus. Under Hypothesis B, a chronic antepartum hypoxic compromise reduces fetal reserve, so that an acute event, including the physiological stress of labour, causes injury it would not otherwise cause.

2. Materials and Methods

2.1. Study Design and Registration

This systematic review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines [11]. The review protocol was registered in the International Prospective Register of Systematic Reviews (PROSPERO) [CRD420261435362] [12]. The review aimed to systematically evaluate the available evidence on the association between maternal diabetes, GDM and PGDM, and hypoxic-ischaemic perinatal injury, with particular focus on acute intrapartum/peripartum evidence, diagnosed neonatal neurological outcome, and chronic antepartum markers of fetal hypoxic compromise. Several elements were defined after the first analysis and are therefore post hoc: the outcome tiers and the exclusion of studies reporting only Tier 3 outcomes, the requirement for a concurrent non-diabetic comparator, the judgement of outcome-definition adequacy, GRADE and SWiM reporting, the vote-counting rules, the classification of each comparison by the direction of its point estimate, the exploratory pooling of umbilical arterial pH and the sensitivity syntheses. They were applied to all records. The protocol was first registered in PROSPERO on 27 June 2026, and the record was amended on 19 October 2026 to describe these post hoc elements.

2.2. Search Strategy

A systematic search of PubMed (NCBI), Embase (Embase.com), and Scopus was run from inception to 16 October 2026. The strategy combined three concept blocks with the Boolean operator AND: (1) diabetes in pregnancy (gestational diabetes, pregnancy in diabetics, diabetes mellitus, hyperglycaemia); (2) pregnancy, fetus, and newborn; and (3) hypoxia-related outcomes and markers, comprising hypoxic-ischaemic or neonatal encephalopathy, asphyxia, fetal hypoxia or distress, neonatal seizures, therapeutic hypothermia, Apgar score, umbilical cord blood gases, acidaemia, base excess or deficit, lactate, cardiotocography and fetal heart rate, Doppler velocimetry, erythropoietin, nucleated red blood cells, polycythaemia, placental pathology, fetal cardiac function, and autopsy. Limits were English language and human studies. Reference lists of included studies were screened to identify additional studies. The search strategies were developed by D.C. and checked against the PRESS 2015 checklist [13] by M.M. before the search was run; because M.M. is a co-author and not an information specialist, this check was not independent of the review team. The complete strategy for each database, with platform, date, limits, and number of records retrieved, is provided in Supplementary Table S1.

2.3. Eligibility Criteria

Studies were included if they met the following criteria: original observational studies (prospective or retrospective cohort, case-control or cross-sectional) in human pregnancies complicated by GDM and/or pregestational diabetes (type 1 or type 2); with a concurrent comparator group of pregnancies without diabetes; reporting at least one Tier 1, Tier 2, or Tier 4 outcome as defined in Section 2.7 (diagnosed or coded neonatal encephalopathy, hypoxic-ischaemic encephalopathy or neonatal seizures, umbilical acid–base status or fetal oxygenation at birth, or a chronic or antepartum marker of fetal hypoxia); and published in English. Studies were excluded if they enrolled a population not meeting this definition, or in which the diabetic subgroup could not be separated; had no non-diabetic comparator, including studies comparing subgroups of women with diabetes only or using historical or published reference values or reference ranges instead of a concurrent comparator group; reported only Tier 3 outcomes (Apgar scores, resuscitation, neonatal intensive care admission, or undefined “asphyxia”); were case series or case reports; were narrative or systematic reviews, editorials, letters, book chapters, conference abstracts, or animal or in vitro studies; or provided insufficient data for extraction. Studies in which all women with diabetes had a coexisting complication that was also present in the comparator group (for example, diabetes with hypertension compared with hypertension alone) met these criteria and were retained, but they were not counted in the synthesis because they do not estimate the effect of diabetes itself (Section 2.8).

2.4. Study Selection

Study selection followed the PRISMA 2020 two-stage screening process [11], as shown in Supplementary Table S2. Titles and abstracts of all records identified by the search were independently screened for relevance by D.C. and S.I.; records clearly not meeting the eligibility criteria were excluded at this stage. Full-text reports of the remaining records were then retrieved and independently assessed against the inclusion and exclusion criteria described in Section 2.3. Disagreements at either stage were to be resolved by discussion between the two reviewers and, where unresolved, by consultation with a third reviewer (G.M.M.). The two reviewers agreed on all records at both stages, so no disagreement required resolution, and the third reviewer was not consulted. After the first analysis, all reports excluded at full text for lacking a Tier 1, 2, or 4 outcome were re-assessed against the full text; reports whose full text could not be obtained were reclassified as not retrieved.

2.5. Quality Assessment

Methodological quality and risk of bias were assessed using the Newcastle–Ottawa Scale (NOS) [14], applied in its cohort-study form (Selection, Comparability, Outcome; maximum 9 stars) or case-control form (Selection, Comparability, Exposure; maximum 9 stars) according to study design. Studies were classified as low risk of bias (7–9 stars), moderate risk of bias (4–6 stars), or high risk of bias (0–3 stars). Quality assessment was performed independently by D.C. and S.I. with disagreements resolved by discussion between the two reviewers and, where unresolved, by consultation with a third reviewer (G.M.M.). One deliberate modification of the NOS was applied consistently: in case-control studies, controls recruited from routine antenatal clinics were treated as community controls. Because the NOS rewards the method of outcome ascertainment rather than the validity of the outcome definition, we added a separate judgement of the adequacy of each study’s hypoxia-related outcome definition. Adequacy was high for a Tier 1 outcome defined by standard criteria (encephalopathy staging, therapeutic hypothermia, neuroimaging, or neuropathology) or a Tier 2 outcome including base deficit, lactate, or oxygen content; moderate for a Tier 1 diagnosis based on an explicit clinical or coded definition, a Tier 2 outcome based on pH alone, or a Tier 4 biomarker measured in a fetal compartment (erythropoietin, nucleated red cells, blinded placental histology); and low for indirect or non-validated markers (Doppler indices, computerised cardiotocography or fetal cardiac indices, placental morphometry, autopsy findings interpreted as hypoxic) or incompletely reported outcomes. Internal validity (NOS category) and outcome-definition adequacy are reported separately. An overall level of concern for the review question was then assigned as the worse of the two, with high adequacy mapped to low concern, moderate adequacy to moderate concern, and low adequacy to high concern [15]. Study design was classified according to how participants were sampled, not according to the label used in the original publication: studies that selected participants by diabetes status and compared outcomes were classified as cohort studies and appraised with the cohort form, and only studies that selected participants by outcome (neonatal seizures) were classified as case-control studies. Four studies measured exposure and outcome at a single time point (cross-sectional design); they were appraised with the cohort form, in which the item “demonstration that the outcome was not present at the start of the study” was not awarded, and the follow-up items were awarded when the outcome was ascertained in all participants at delivery. Certainty of evidence for each outcome tier, diabetes population, and chronic marker was rated with the GRADE approach for observational evidence [16]. Certainty was not rated up for a large effect when it had been rated down for any domain. Domain-level reasons for every rating are reported in Supplementary Table S5. Selective outcome reporting was assessed by comparing the outcomes listed in each study’s methods with those reported in its results; funnel-plot methods were not applied because no single comparison included 10 or more studies reporting a common effect measure.

2.6. Data Extraction

Data were extracted using a standardised database including study identifier, author and year, country/setting, study design, diabetes type (GDM, PGDM, or mixed population), sample size, diagnostic criteria for diabetes, maternal glycaemic control or treatment, presence and type of comparator group, intrapartum monitoring findings, acute peripartum biochemical and clinical status (umbilical arterial pH, base excess, lactate, Apgar scores, need for resuscitation), diagnosis and staging of hypoxic-ischaemic encephalopathy, use of therapeutic hypothermia, neuroimaging findings, neurodevelopmental follow-up, antepartum/chronic hypoxia markers (Doppler velocimetry, placental histology, hematologic and erythropoietin-based markers), authors’ own interpretation of the relative contribution of acute versus chronic mechanisms, and study limitations. Extraction was performed independently by D.C. and S.I. with discrepancies resolved by discussion. Extracted outcomes were classified into the outcome tiers described in Section 2.7. Each study was also mapped to the working hypothesis it could inform: Hypothesis A (injury caused by an acute intrapartum event in a previously uncompromised fetus) when it reported intrapartum monitoring or acute status at birth and Hypothesis B (chronic antepartum compromise reducing fetal reserve, with a superimposed acute event) when it reported a chronic or antepartum marker; studies reporting both were mapped to both, and studies reporting only a Tier 1 outcome were mapped to neither, because they do not describe the timing of compromise. Diagnostic criteria, setting, glycaemic control, and the mapping of each study are reported in Supplementary Table S4.

2.7. Outcome Classification

Because the included studies reported hypoxia-related outcomes of very different clinical meaning, every outcome was assigned to one of four tiers, and outcomes from different tiers were never combined. Tier 1, diagnosed or coded neonatal encephalopathy, hypoxic-ischaemic encephalopathy or neonatal seizures: clinical encephalopathy with documented staging (Sarnat or Thompson), treatment with therapeutic hypothermia, neuroimaging, neuropathological confirmation, or an explicitly defined or coded diagnosis of hypoxic-ischaemic encephalopathy (HIE), neonatal encephalopathy, or neonatal seizures. Because coded diagnoses and birth-certificate seizure fields do not establish a hypoxic-ischaemic aetiology, and no included study reported encephalopathy staging, Tier 1 outcomes are described as diagnosed or coded rather than confirmed injury. Tier 2, biochemical markers of acute hypoxia at birth: umbilical arterial pH, base excess or lactate, reported as continuous values or at recognised thresholds (e.g., pH < 7.00 or < 7.10; base excess ≤ −12 mmol/L) [17], and umbilical oxygen tension, saturation, or content. Tier 3, non-specific clinical surrogates: Apgar scores, resuscitation or respiratory support, neonatal intensive care admission, and composite “asphyxia” outcomes that were undefined or derived from Apgar thresholds; composite outcomes combining a Tier 2 criterion with an Apgar criterion were classified as Tier 3, and the Tier 2 components were used where they were reported separately. Tier 4, chronic or antepartum markers of fetal hypoxic stress: erythropoietin, nucleated red blood cells, polycythaemia and other indices of increased fetal erythropoiesis (hepatic erythropoiesis, erythrocyte zinc protoporphyrin), placental histology and morphometry, Doppler velocimetry, antepartum fetal heart rate patterns and fetal cardiac function, and stillbirth autopsy findings. Intrapartum fetal heart rate patterns and fetal scalp sampling were not classified as Tier 4, because they describe compromise during labour; they were reported descriptively as intrapartum monitoring and mapped to Hypothesis A. Studies reporting only Tier 3 outcomes were not eligible (Section 2.3), and Tier 3 outcomes reported by included studies were not synthesised. Tier 1 outcomes were interpreted as the closest available proxy for hypoxic-ischaemic brain injury; Tier 2 outcomes were interpreted as evidence of acute compromise at birth, and Tier 4 markers as indicators of chronic fetal hypoxic stress only.

2.8. Data Synthesis

Within each outcome tier, definitions, thresholds, comparators, and diabetes types differed; few studies reported the effect estimates and dispersion statistics required for pooling, and a pooled estimate across different outcome definitions would have combined distinct constructs. Synthesis was therefore conducted without meta-analysis and is reported following the Synthesis Without Meta-analysis (SWiM) guideline [18]. Studies were grouped by outcome tier (Section 2.7) and by diabetes population (GDM only; pregestational diabetes only; both; or unspecified), and studies reporting GDM and pregestational diabetes separately contributed to each population. Findings were summarised by vote counting based on the direction of effect and displayed in an effect-direction plot [19]. The direction of each comparison was classified by its point estimate, irrespective of statistical significance: for Tier 1, the adjusted estimate where reported; for Tier 2, umbilical arterial pH or, if not reported, the main acid–base measure; and for Tier 4, the main marker reported. A comparison was classified as showing no difference (◄►) only when the values in the two groups were identical or the study reported no numerical data; statistical significance is reported alongside direction. The following vote-counting rules were defined post hoc: (i) comparisons with fewer than five events in total were classified as uninformative and did not contribute a vote; (ii) reports from the same research group with documented or possible sample overlap were counted once in the primary synthesis; (iii) studies whose comparator group had another pregnancy complication (pre-eclampsia or hypertension) were described but not counted; and (iv) placental findings that the authors interpreted as compensatory adaptation were coded separately and not counted as adverse. An exploratory random-effects summary of the mean difference in umbilical arterial pH in GDM was restricted to studies reporting the mean and standard deviation against a non-diabetic comparator. Between-study variance was estimated with the DerSimonian–Laird method and, because only three studies were available, the 95% confidence interval was computed with the Hartung–Knapp–Sidik–Jonkman method, and a 95% prediction interval is reported [20,21]. Where a study reported separate GDM subgroups (diet- and insulin-treated), they were combined into a single group with the standard formula for combining means and standard deviations. In a sensitivity analysis, two studies reporting medians and interquartile ranges were added after conversion to means and standard deviations [22]. Post hoc sensitivity syntheses (a) were restricted to studies rated low or moderate overall concern, (b) counted overlapping reports separately, (c) for Tier 2 outcomes in GDM, were restricted to studies of elective caesarean or uncomplicated delivery, and (d) for Tier 4 outcomes, excluded studies whose comparators were autopsied infants who died of other conditions or fetuses selected as non-hypoxic.

3. Results

3.1. Search Results

The database search identified 14,959 records (PubMed 5356; Embase 2734; Scopus 6869; Supplementary Table S1). After removal of 4174 duplicates, 10,785 records were screened by title and abstract, and 10,627 were excluded. Of the 158 reports sought for retrieval, 25 could not be retrieved. Of the 133 reports assessed for eligibility, 59 were excluded (Supplementary Table S3): 36 had no non-diabetic comparator, nine did not report a Tier 1, 2, or 4 outcome, six were case series or case reports, four had an ineligible population or did not report the diabetic group separately, two were duplicate or overlapping reports of included studies, and two were a letter and a book chapter. Seventy-four studies were included (Figure 1).

Figure 1. Flowchart of the included studies.

3.2. Risk of Bias Assessment

Newcastle–Ottawa Scale scores ranged from 4 to 9 stars (Table 1): 28 studies were rated low risk of bias (7–9 stars) and 46 moderate (4–6 stars); none scored in the high-risk range. Stars were most often withheld for restricted or convenience samples (selection), for the absence of adjustment for confounders, which applied to most biomarker, Doppler, and placental studies (comparability), and for diabetes ascertainment by administrative codes or unstated criteria. Outcome-definition adequacy was high in 17 studies, moderate in 35, and low in 22. Combining the two judgements, overall concern for the review question was low in eight studies [23,24,25,26,27,28,29,30], moderate in 44, and high in 22. High concern was concentrated among Tier 4 studies relying on Doppler indices, computerised cardiotocography or fetal cardiac indices, placental morphometry, or autopsy findings, in which a hypoxic mechanism was inferred rather than measured. Two clusters of overlapping populations were identified: four studies of fetal cardiac function from one South African group [31,32,33,34] and three placental stereology studies from one Nottingham–Aarhus collaboration [35,36,37]; each cluster was counted as a single evidence source in the primary synthesis. Three studies from the Helsinki University Hospital diabetes and pregnancy programme [38,39,40] enrolled women in different periods and reported different outcomes and were counted separately. Regarding selective outcome reporting, one study reported cord blood gases for only 20 of 42 infants, in Supplementary Materials [41]; in the remaining studies, the hypoxia-related outcomes described in the methods were reported in the results.

Table 1. Risk of bias (Newcastle–Ottawa Scale), adequacy of the hypoxia-related outcome definition, and overall concern for the review question.

Study	Design	Selection (S1–S4)	Comparability	Outcome/Exposure (O1–O3)	NOS Stars	NOS Risk of Bias	Outcome-Definition Adequacy	Overall Concern
Gestational diabetes only (n = 30)
Danciu [42]	Prospective cohort (comparative)	– ★ ★ ★	–	★ ★ ★	6	Moderate	Low	High
Emiralioglu [43]	Prospective cohort	★ ★ ★ ★	★★	★ ★ ★	9	Low	Low	High
Kayaalp [44]	Prospective cohort	– ★ ★ ★	–	★ ★ ★	6	Moderate	High	Moderate
Kömmel [45]	Prospective cohort	– ★ ★ ★	★	★ ★ ★	7	Low	Low	High
Szablewska [46]	Prospective-retrospective cohort	– ★ ★ ★	–	★ ★ ★	6	Moderate	High	Moderate
Golshahi [26]	Retrospective cohort (elective CS)	– ★ ★ ★	★	★ ★ ★	7	Low	High	Low
Gul Kara [47]	Prospective cohort	– ★ ★ ★	–	★ ★ ★	6	Moderate	Moderate	Moderate
Zhang [41]	Prospective cohort (omics)	– ★ ★ ★	–	★ ★ –	5	Moderate	Low	High
Boskabadi [48]	Cohort (preterm infants)	– ★ – ★	–	★ ★ ★	5	Moderate	Low	High
Mor [49]	Retrospective cohort	– ★ – ★	★	★ ★ ★	6	Moderate	Moderate	Moderate
Sappler [27]	Matched cohort (very preterm)	– ★ ★ ★	★	★ ★ ★	7	Low	High	Low
Zhang [50]	Retrospective cohort (preterm)	– ★ ★ ★	–	– ★ ★	5	Moderate	Low	High
Dasgupta [51]	Prospective cohort	– ★ ★ ★	–	★ ★ ★	6	Moderate	Moderate	Moderate
Sutreja [52]	Cross-sectional comparative	– ★ ★ –	–	★ ★ ★	5	Moderate	Low	High
Bhorat [31]	Prospective matched cohort	★ ★ ★ ★	★★	★ ★ –	8	Low	Moderate	Moderate
Bianchi [28]	Prospective cohort (elective CS)	– ★ ★ ★	★	★ ★ ★	7	Low	High	Low
Chen [53]	Retrospective cohort (SGA infants only)	– ★ ★ ★	–	★ ★ ★	6	Moderate	Moderate	Moderate
Mohd Idris [54]	Prospective cohort (laboratory)	– ★ ★ ★	–	★ ★ ★	6	Moderate	Moderate	Moderate
Tarvonen [23]	Retrospective cohort	★ ★ ★ ★	★	★ ★ ★	8	Low	High	Low
Bhorat [32]	Prospective matched cross-sectional	– ★ – –	★	★ ★ ★	5	Moderate	Low	High
Bhorat [33]	Prospective matched cross-sectional	– ★ – –	★	★ ★ ★	5	Moderate	Low	High
Kovo [55]	Retrospective cohort	– ★ ★ ★	–	★ ★ ★	6	Moderate	Moderate	Moderate
Beneventi [29]	Retrospective cohort (uncomplicated vaginal births)	– ★ ★ ★	★★	★ ★ ★	8	Low	High	Low
Bhorat [34]	Prospective matched cohort	– ★ – ★	★	★ ★ ★	6	Moderate	Low	High
Taricco [25]	Prospective cohort (elective CS)	– ★ ★ ★	★★	★ ★ ★	8	Low	High	Low
Daskalakis [56]	Prospective cohort (placental pathology)	– ★ ★ ★	–	★ ★ ★	6	Moderate	Moderate	Moderate
Madazli [57]	Prospective matched cohort	– ★ ★ ★	★	★ ★ ★	7	Low	Moderate	Moderate
Sarkar [58]	Prospective cohort	– ★ ★ ★	–	★ ★ ★	6	Moderate	Moderate	Moderate
Yeruchimovich [59]	Prospective cohort	– ★ ★ ★	★	★ ★ ★	7	Low	Moderate	Moderate
Drexel [60]	Prospective matched cohort	– ★ ★ ★	★	★ ★ –	6	Moderate	Moderate	Moderate
Pregestational diabetes only (n = 12)
Rizvi [61]	Retrospective matched cohort	★ ★ ★ ★	★★	★ ★ ★	9	Low	Moderate	Moderate
Agaoglu [62]	Prospective cohort	– ★ ★ ★	–	★ ★ ★	6	Moderate	High	Moderate
Cnattingius [63]	Population-based cohort	★ ★ ★ ★	★★	★ ★ ★	9	Low	Moderate	Moderate
Verhaeghe [64]	Prospective cohort	– ★ ★ ★	–	★ ★ ★	6	Moderate	High	Moderate
Verner [65]	Prospective cohort	– ★ ★ ★	–	★ ★ ★	6	Moderate	Moderate	Moderate
Jauniaux [66]	Cohort (placental morphometry)	– ★ ★ ★	–	★ ★ ★	6	Moderate	Low	High
Ruozi-Berretta [67]	Prospective cohort (elective CS)	– ★ – ★	–	★ ★ ★	5	Moderate	Low	High
Teramo [38]	Prospective cohort	– ★ ★ ★	★	★ ★ ★	7	Low	Moderate	Moderate
Mayhew [36]	Cohort (placental stereology)	– ★ ★ ★	–	★ ★ ★	6	Moderate	Low	High
Ramanathan [68]	Prospective comparative cohort (elective caesarean, epidural)	– ★ ★ ★	–	★ ★ ★	6	Moderate	High	Moderate
Widness [40]	Prospective cohort (elective CS)	– ★ ★ ★	–	★ ★ ★	6	Moderate	Moderate	Moderate
Mimouni [69]	Prospective pair-matched cohort	– ★ ★ ★	★★	★ ★ ★	8	Low	Moderate	Moderate
Gestational and pregestational, or type not specified (n = 32)
Seyhanli [70]	Prospective cohort	– ★ ★ ★	–	★ ★ ★	6	Moderate	Low	High
Kopteeva [71]	Prospective cohort	– ★ ★ ★	–	★ ★ ★	6	Moderate	High	Moderate
Karkia [4]	Prospective cohort	★ ★ ★ ★	★★	★ ★ ★	9	Low	Moderate	Moderate
Liang [72]	Nested case-control (propensity-matched)	★ ★ – ★	★★	– ★ ★	7	Low	Moderate	Moderate
Bandoli [24]	Population-based cohort	★ ★ – ★	★★	★ ★ ★	8	Low	High	Low
Lynch [73]	Retrospective cohort of autopsied stillbirths	– ★ ★ ★	★	★ ★ ★	7	Low	Low	High
Mahon [74]	Prospective cohort (term AGA)	– ★ – ★	–	★ ★ ★	5	Moderate	Moderate	Moderate
Page [75]	Cohort of stillbirths (secondary analysis)	★ ★ ★ ★	★	★ ★ ★	8	Low	Low	High
Beyerlein [76]	Population-based cohort (registry)	★ ★ – ★	–	★ ★ –	5	Moderate	Moderate	Moderate
Castelijn [77]	Retrospective case-cohort	★ ★ ★ ★	★★	★ ★ –	8	Low	Moderate	Moderate
Zaigham [30]	Retrospective cohort	★ ★ ★ ★	–	★ ★ ★	7	Low	High	Low
Stanek [78]	Retrospective cohort (placental pathology)	– ★ – ★	–	★ ★ ★	5	Moderate	Moderate	Moderate
Stanescu [79]	Retrospective cohort	– ★ – ★	–	★ ★ ★	5	Moderate	High	Moderate
Edwards [80]	Matched cohort of stillbirths (post-mortem)	★ ★ ★ ★	★	★ ★ ★	8	Low	Low	High
Reif [81]	Retrospective cohort (labours with suspected fetal distress)	– ★ ★ ★	★	★ ★ ★	7	Low	Moderate	Moderate
Potti [82]	Population-based cohort (linked records)	★ ★ – ★	★★	★ ★ ★	8	Low	Moderate	Moderate
Cetin [83]	Prospective cohort	– ★ ★ ★	–	★ ★ ★	6	Moderate	Moderate	Moderate
Namavar Jahromi [84]	Prospective cohort (caesarean births)	– ★ ★ ★	–	★ ★ ★	6	Moderate	High	Moderate
Russell [85]	Retrospective cohort of autopsied stillbirths	– ★ ★ ★	–	★ ★ ★	6	Moderate	Low	High
Hall [86]	Population-based case-control (birth certificates)	– ★ ★ ★	★	★ ★ ★	7	Low	Moderate	Moderate
Lott [87]	Prospective cohort (consecutive deliveries)	– ★ ★ ★	–	★ ★ ★	6	Moderate	Moderate	Moderate
Loukovaara [39]	Prospective cohort	– ★ ★ ★	–	★ ★ ★	6	Moderate	Moderate	Moderate
Tan [88]	Matched cohort	– ★ ★ ★	★	★ ★ ★	7	Low	Moderate	Moderate
Vatansever [89]	Prospective cohort (high-risk neonates)	– ★ – ★	–	★ ★ ★	5	Moderate	Moderate	Moderate
Mayhew [35]	Cohort (placental stereology)	– – – ★	–	★ ★ ★	4	Moderate	Low	High
Mayhew [37]	Cohort (placental stereology)	– ★ ★ ★	–	★ ★ ★	6	Moderate	Low	High
Olofsson [90]	Prospective matched cohort (planned vaginal delivery)	– ★ ★ ★	★★	★ ★ –	7	Low	Moderate	Moderate
Shannon [91]	Prospective cohort	– ★ ★ ★	–	★ ★ ★	6	Moderate	Moderate	Moderate
Singer [92]	Retrospective autopsy comparison	– – ★ ★	–	★ ★ ★	5	Moderate	Low	High
Widness [93]	Prospective cohort	– ★ – ★	–	★ ★ ★	5	Moderate	Moderate	Moderate
Fadel [94]	Cross-sectional (amniocentesis)	– ★ – –	–	★ ★ ★	4	Moderate	Low	High
Datta [95]	Prospective comparative cohort (elective caesarean)	– ★ ★ ★	–	★ ★ ★	6	Moderate	High	Moderate

Abbreviations: NOS, Newcastle–Ottawa Scale. In the Selection and Outcome/exposure columns, each symbol corresponds to one NOS item, in the order listed in the column header: ★, criterion met (one star); –, criterion not met. For example, ★ ★ ★ ★ indicates that all four selection criteria (S1–S4) were met (4 stars), ★ ★ ★ that all three outcome or exposure criteria (O1–O3 or E1–E3) were met (3 stars), and – ★ ★ ★ that S1 was not met and S2–S4 were met (3 stars). For cohort designs, S1, representativeness of the exposed cohort; S2, selection of the non-exposed cohort; S3, ascertainment of exposure; S4, demonstration that the outcome was not present at the start of the study; O1, assessment of outcome; O2, follow-up long enough for outcomes to occur; O3, adequacy of follow-up. For case-control designs, S1, adequacy of case definition; S2, representativeness of cases; S3, selection of controls; S4, definition of controls; and the third group of criteria refers to exposure: E1, ascertainment of exposure; E2, same method of ascertainment for cases and controls; E3, non-response rate. Comparability: ★, one star (comparability for the most important factor); ★★, both comparability stars (most important factor and an additional factor); –, not met. Maximum 9 stars. NOS risk of bias: 7–9 stars, low; 4–6, moderate; 0–3, high. For cross-sectional designs, S4 (outcome not present at start) was not awarded (Section 2.5). Outcome-definition adequacy: high, Tier 1 outcome defined by staging, therapeutic hypothermia, neuroimaging or neuropathology, or Tier 2 outcome including base deficit, lactate or oxygen content; moderate, Tier 1 diagnosis by explicit clinical or coded definition, Tier 2 outcome based on pH alone, or Tier 4 biomarker measured in a fetal compartment; low, indirect or non-validated markers or incompletely reported outcomes. Overall concern is the worse of the NOS risk of bias and the adequacy rating (high adequacy, low concern; moderate, moderate; low, high).

Stars were withheld as follows: S1, restricted samples (preterm-only, small-for-gestational-age-only, stillbirth-only, elective caesarean only) or convenience samples; S2, comparator drawn from a different source population; S3, diabetes ascertained by administrative codes or diagnostic criteria not stated; comparability, unadjusted, or unmatched comparison; O1, outcome not defined or not measured in all participants; O3, incomplete outcome data not accounted for.

3.3. Characteristics of Included Studies

The 74 included studies were published between 1977 and 2026 (46 since 2010) and came from 22 countries, most frequently the United States (n = 18), Turkey (n = 8), Finland, Italy, South Africa and the United Kingdom (n = 4 each), and Austria, China, Denmark (three UK–Danish collaborations), Iran, Israel, and Sweden (n = 3 each) (Table 2). Thirty studies enrolled GDM only, 12 pregestational diabetes only, and 32 both types or did not specify the type; type 2 diabetes was reported separately in only three studies [62,63,77]. Sample sizes ranged from 22 pregnancies to approximately 3 million births. Designs, classified by sampling (Section 2.5), comprised prospective and retrospective cohorts, population-based registry cohorts, two case-control studies of neonatal seizures, one case-cohort study, five studies restricted to stillbirths or autopsies, and four cross-sectional studies. Twelve studies reported a Tier 1 outcome, 30 a Tier 2 outcome, and 44 a Tier 4 outcome; 12 reported outcomes in more than one tier. Two further studies reported intrapartum fetal heart rate monitoring, which was not classified as a Tier 4 outcome (Section 2.7). No study reported Sarnat or Thompson staging. Diagnostic criteria, setting, and maternal glycaemic control or treatment are given for each study in Supplementary Table S4; a measure of glycaemic control or treatment was reported in about half of the studies.

Table 2. Characteristics of included studies.

Study	Year	Country	Design	Diabetes Type	Sample Size (Diabetic; Comparator)	Comparator	Outcome Tier(s)	Hypoxia-Related Outcomes Assessed
Gestational diabetes only (n = 30)
Danciu [42]	2026	Romania	Prospective cohort (comparative)	GDM	488 (146 GDM; 87 low-risk controls)	Low-risk pregnancies	Tier 4	Antepartum computerised CTG (STV, LTV)
Emiralioglu [43]	2026	Turkey	Prospective cohort	GDM	175 (83 GDM; 92 controls)	Non-GDM	Tier 4	Fetal MCA/UA/DV Doppler incl. CPR, MCA diastolic deceleration area
Kayaalp [44]	2026	Turkey	Prospective cohort	GDM	117 (64 GDM; 53 controls)	Non-GDM	Tier 2, Tier 4	CPR, CPUR, UA-PI; cord arterial pH and BE
Kömmel [45]	2026	Germany	Prospective cohort	GDM	172 (56 GDM; 116 controls)	Healthy controls	Tier 4	Fetal ECG heart time intervals and HRV; cCTG
Szablewska [46]	2026	Poland	Prospective-retrospective cohort	GDM	191 (44 GDM; 147 controls)	Non-GDM	Tier 2	Umbilical arterial pH, pCO2, pO2, BE, HCO3
Golshahi [26]	2025	Iran	Retrospective cohort (elective CS)	GDM	430 (87 GDM)	Non-GDM, elective CS	Tier 2	Umbilical arterial pH, BD, pCO2, pO2, HCO3
Gul Kara [47]	2025	Turkey	Prospective cohort	GDM	108 (51 GDM; 57 controls)	Normotensive non-GDM	Tier 2	Cord pH; cord sFlt-1/PlGF
Zhang [41]	2025	China	Prospective cohort (omics)	GDM (±PE)	42 (12 GDM; 10 GDM+PE; 10 PE; 10 controls)	Uncomplicated pregnancies	Tier 2	Cord blood gases (20/42 infants)
Boskabadi [48]	2024	Iran	Cohort (preterm infants)	GDM	295 preterm (99 IDM; 196 controls)	Preterm infants of healthy mothers	Tier 2, Tier 4	Venous blood gas pH; NRBC/100 WBC
Mor [49]	2024	Israel	Retrospective cohort	GDM (±PE)	1750 (1532 GDM-exposed)	Isolated pre-eclampsia	Tier 4	Placental histology (Amsterdam criteria)
Sappler [27]	2023	Austria	Matched cohort (very preterm)	GDM	141 (47 GDM; 94 matched controls)	GA/sex/scanner-matched preterm infants	Tier 1	Term-equivalent MRI brain injury (Kidokoro score); DTI
Zhang [50]	2023	China	Retrospective cohort (preterm)	GDM	640 preterm (217 GDM)	Non-GDM preterm	Tier 1	HIE (criteria not specified)
Dasgupta [51]	2022	India	Prospective cohort	GDM	84 (42 GDM; 42 controls)	Screen-negative pregnancies	Tier 4	Blinded placental histology (two pathologists)
Sutreja [52]	2022	India	Cross-sectional comparative	GDM	100 placentas (50 GDM; 50 normal)	Uncomplicated pregnancies	Tier 4	Placental histology
Bhorat [31]	2021	South Africa	Prospective matched cohort	GDM (poorly controlled)	88 (44 GDM; 44 controls)	Matched controls	Tier 2, Tier 4	Mod-MPI, Doppler, nRBC, cord pH < 7.2
Bianchi [28]	2021	Italy	Prospective cohort (elective CS)	GDM (overweight/obese)	334 (34 GDM; 300 non-GDM)	Normal-weight and non-GDM women	Tier 2	Umbilical arterial/venous pO2, O2 saturation and content, pH, lactate
Chen [53]	2021	China	Retrospective cohort (SGA infants only)	GDM	1981 SGA infants (383 GDM)	SGA infants of non-GDM mothers	Tier 1	HIE (clinical criteria incl. Apgar ≤ 5 at 5 min and encephalopathy)
Mohd Idris [54]	2021	Malaysia	Prospective cohort (laboratory)	GDM	80 (42 GDM; 38 controls)	Normal pregnancies	Tier 4	Umbilical cord plasma EPO
Tarvonen [23]	2021	Finland	Retrospective cohort	GDM	5150 (624 GDM)	Normal OGTT; no OGTT	Tier 1, Tier 2 (intrapartum CTG reported separately)	Intrapartum CTG ZigZag pattern, late decelerations; UA pH/BE; NE
Bhorat [32]	2019	South Africa	Prospective matched cross-sectional	GDM (well-controlled)	108 (54 GDM; 54 controls)	Matched normal pregnancies	Tier 4	Fetal MPI
Bhorat [33]	2018	South Africa	Prospective matched cross-sectional	Gestational impaired glucose tolerance (diet)	64 (32 GIGT; 32 controls)	Matched normal pregnancies	Tier 4	Fetal MPI
Kovo [55]	2016	Israel	Retrospective cohort	GDM (±hypertensive disorders)	233 (192 GDM; 41 non-GDM hypertensive)	Non-GDM hypertensive pregnancies	Tier 4	Placental maternal/fetal vascular supply lesions
Beneventi [29]	2014	Italy	Retrospective cohort (uncomplicated vaginal births)	GDM	426 (142 GDM; 284 controls)	Non-GDM uncomplicated vaginal births	Tier 2	Cord arterial pH, BD, lactate
Bhorat [34]	2014	South Africa	Prospective matched cohort	GDM (poorly controlled, insulin)	58 (29 GDM; 29 controls)	Matched normal pregnancies	Tier 4	Fetal Mod-MPI, E/A ratio (cardiac Doppler)
Taricco [25]	2009	Italy	Prospective cohort (elective CS)	GDM	75 (38 GDM; 37 controls)	Normal pregnancies	Tier 2, Tier 4	UV/UA O2 saturation and content, lactate; placental histology
Daskalakis [56]	2008	Greece	Prospective cohort (placental pathology)	GDM	80 (40 GDM; 40 controls)	Consecutive normal pregnancies	Tier 4	Blinded placental histology incl. NRBC, villous immaturity, chorangiosis
Madazli [57]	2008	Turkey	Prospective matched cohort	GDM	44 (22 GDM; 22 controls)	OGTT-negative women matched for GA, age and BMI	Tier 4	Cord EPO, NRBC; blinded placental histology
Sarkar [58]	2003	USA	Prospective cohort	GDM A1 (diet)	160 (92 GDM; 68 controls)	Non-diabetic	Tier 1	HIE (clinical)
Yeruchimovich [59]	2000	Israel	Prospective cohort	GDM	70 (40 GDM [20 LGA, 20 AGA]; 30 controls)	AGA infants of non-diabetic women	Tier 4	Absolute NRBC count within 12 h; haematocrit
Drexel [60]	1988	Austria	Prospective matched cohort	GDM	204 (102 GDM; 102 matched controls)	Matched non-diabetic pregnancies	Tier 2	Umbilical arterial pH < 7.20 and <7.10
Pregestational diabetes only (n = 12)
Rizvi [61]	2025	USA	Retrospective matched cohort	T1D	322 (161 T1D; 161 controls)	Matched healthy controls	Tier 1	HIE (named outcome); “hypoxia”
Agaoglu [62]	2024	Turkey	Prospective cohort	PGDM (T1D, T2D)	195 (65 PGDM; 130 controls)	Low-risk pregnancies	Tier 2, Tier 4	CPR, CPUR; cord arterial pH and BE
Cnattingius [63]	2017	Sweden	Population-based cohort	T1D; T2D	1,343,751 (5941 T1D; 711 T2D)	No diabetes	Tier 1	HIE or neonatal seizures (ICD-10)
Verhaeghe [64]	2011	Belgium	Prospective cohort	PGDM (T1D, T2D)	141 (70 PGDM; 71 controls)	Newborns of non-diabetic mothers	Tier 2	Umbilical arterial pH, pO2, O2 saturation, pCO2, base excess
Verner [65]	2007	United Kingdom	Prospective cohort	T1D	97 (49 T1D; 48 controls)	Non-diabetic	Tier 2, Tier 4	Cord pH; cord EPO; iron indices
Jauniaux [66]	2006	United Kingdom	Cohort (placental morphometry)	T1D	22 (12 T1D; 10 controls)	Uncomplicated elective CS	Tier 4	Villous morphometry; villous membrane diffusing capacity
Ruozi-Berretta [67]	2004	Italy	Prospective cohort (elective CS)	PGDM	73 (27 PGDM; 46 controls)	Normal pregnancies	Tier 2, Tier 4	Antepartum cCTG (STV, accelerations); umbilical blood gases
Teramo [38]	2004	Finland	Prospective cohort	T1D	175 (156 T1D; 19 controls)	Healthy controls at CS	Tier 4	Amniotic fluid EPO
Mayhew [36]	1993	Denmark (UK–Danish collaboration)	Cohort (placental stereology)	Diabetes (White class D)	Control and diabetic placentas	Uncomplicated pregnancies	Tier 4	Villous and capillary morphometry
Ramanathan [68]	1991	USA	Prospective comparative cohort (elective caesarean, epidural)	T1D	50 (20 IDDM; 15 euglycaemic and 15 iatrogenically hyperglycaemic controls)	Healthy euglycaemic women (group 3)	Tier 2	Umbilical arterial and venous pH, pO2, pCO2, base excess, lactate
Widness [40]	1990	Finland	Prospective cohort (elective CS)	T1D	67 (44 T1D; 23 controls)	Non-diabetic	Tier 4	Umbilical venous EPO
Mimouni [69]	1986	USA	Prospective pair-matched cohort	PGDM (White B–R)	68 (34 IDM; 34 matched controls)	Infants of non-diabetic mothers matched for GA, mode of delivery, Apgar group, sampling site/time and cord clamping	Tier 4	Polycythaemia (venous haematocrit ≥ 65% at 2–4 h); NRBC/100 WBC
Gestational and pregestational, or type not specified (n = 32)
Seyhanli [70]	2026	Turkey	Prospective cohort	PGDM; GDM (diet/insulin)	165 (30 PGDM; 60 GDM; 75 controls)	Non-diabetic controls	Tier 2, Tier 4	Aortic isthmus, UA, MCA, UtA Doppler; cord pH
Kopteeva [71]	2025	Russia	Prospective cohort	PGDM; GDM	188 (86 PGDM; 44 GDM; 58 controls)	Non-diabetic controls	Tier 2, Tier 4	Fetal liver/ductus venosus volumetric flow; fetal hypoxia (cord pH < 7.10, lactate > 6.1 or BE < −9)
Karkia [4]	2023	United Kingdom	Prospective cohort	PGDM; GDM	53,649 (509 PGDM; 2089 GDM)	No diabetes	Tier 1	HIE (abnormal neurology with Apgar < 5 or cord pH < 7)
Liang [72]	2023	USA (NVSS data)	Nested case-control (propensity-matched)	PGDM; GDM	33,370 (6674 cases; 26,696 controls)	Neonates without seizures	Tier 1	Neonatal seizures (birth-certificate field)
Bandoli [24]	2022	USA	Population-based cohort	Pre-existing diabetes; GDM	2,994,892 (GDM ~11%)	No diabetes	Tier 1	Neonatal encephalopathy (ICD); NE with therapeutic hypothermia
Lynch [73]	2022	USA	Retrospective cohort of autopsied stillbirths	PGDM; GDM	75 stillbirths (15 diabetic; 60 non-diabetic)	Non-diabetic stillbirths	Tier 4	Fetal autopsy: cardiomyopathy, organ weights, causes of death
Mahon [74]	2021	USA	Prospective cohort (term AGA)	Diabetes (PGDM/GDM)	52 (25 diabetic; 27 controls)	Non-diabetic AGA newborns	Tier 2	Cord arterial pH, pO2, catecholamines
Page [75]	2020	USA	Cohort of stillbirths (secondary analysis)	PGDM; GDM	455 stillbirths (46 diabetic)	Non-diabetic stillbirths	Tier 4	Standardised cause of death incl. placental disease; postmortem and placental pathology
Beyerlein [76]	2018	Germany	Population-based cohort (registry)	PGDM; GDM	1,716,170 deliveries (2001–2016)	No diabetes	Tier 2	Umbilical cord pH < 7.1
Castelijn [77]	2018	Netherlands	Retrospective case-cohort	T1D; T2D; GDM	15,739 (117 T1D; 59 T2D; 303 GDM)	Non-diabetic controls	Tier 2	UA pH < 7.20 and <7.10; fetal distress
Zaigham [30]	2018	Sweden	Retrospective cohort	PGDM; GDM	17,358 (574 diabetic)	Non-diabetic	Tier 2	Cord arterial acidaemia (pH < mean-2SD), metabolic acidosis, lactate
Stanek [78]	2017	USA	Retrospective cohort (placental pathology)	Maternal diabetes (PGDM/GDM)	5248 placentas (287 diabetic)	Remaining placentas	Tier 4	Placental histological patterns of chronic hypoxic injury
Stanescu [79]	2014	Romania	Retrospective cohort	Diabetes (type NS)	240 (120 diabetic; 120 controls)	Uncomplicated pregnancies	Tier 2	Cord pH, BE, HCO3
Edwards [80]	2013	United Kingdom	Matched cohort of stillbirths (post-mortem)	Diabetes (PGDM/GDM)	46 stillbirths (23 matched pairs)	Matched non-diabetic stillbirths	Tier 4	Post-mortem findings (thymic starry-sky involution, placental weight)
Reif [81]	2013	Austria	Retrospective cohort (labours with suspected fetal distress)	GDM; T1D	171 (57 diabetic; 114 controls)	Non-diabetic with non-reassuring FHR	Tier 2	Cord arterial pH; fetal scalp pH
Potti [82]	2012	USA	Population-based cohort (linked records)	Diabetes (type NS)	~1,000,000 (4.79% diabetes)	No diabetes or hypertension	Tier 1	Neonatal seizures (birth certificate)
Cetin [83]	2011	Turkey	Prospective cohort	PGDM; GDM	47 (27 diabetic; 20 controls)	Normoglycaemic controls	Tier 4	Cord plasma EPO; haematocrit
Namavar Jahromi [84]	2011	Iran	Prospective cohort (caesarean births)	Insulin-requiring diabetes	478 (68 diabetic; 410 controls)	Healthy pregnancies	Tier 2, Tier 4	UA pH, pCO2, BE; NRBC
Russell [85]	2008	Ireland	Retrospective cohort of autopsied stillbirths	T1D; GDM	155 term stillbirths (30 diabetic; 18 LGA and 107 AGA non-diabetic)	Unexplained non-diabetic stillbirths	Tier 4	Autopsy cardiomegaly, heart and brain weight, ventricular wall thickness
Hall [86]	2006	USA	Population-based case-control (birth certificates)	PGDM; GDM	6234 term neonates (567 cases; 5667 controls)	Term neonates without seizures	Tier 1	Neonatal seizures within 72 h (birth-certificate record)
Lott [87]	2005	USA	Prospective cohort (consecutive deliveries)	Insulin-treated (type NS)	118 (18 FDM; 100 controls; 36 IUGR not used)	Non-hypoxic control fetuses	Tier 4	Cord plasma EPO; RBC zinc protoporphyrin/heme; ferritin
Loukovaara [39]	2005	Finland	Prospective cohort	T1D; GDM (insulin)	157 (67 T1D; 28 GDM; 62 controls)	Normal pregnancies	Tier 2	Umbilical arterial pH, pO2, pCO2 (amniotic fluid EPO in diabetic pregnancies only)
Tan [88]	2005	Malaysia	Matched cohort	Diabetes (PGDM/GDM)	100 (50 diabetic; 50 controls)	Matched non-diabetic	Tier 2, Tier 4	UA resistance index; umbilical arterial acidosis
Vatansever [89]	2002	Turkey	Prospective cohort (high-risk neonates)	T1D; GDM	106 (12 IDM; 37 healthy controls)	Healthy term infants	Tier 4	Neonatal NRBC count; serum EPO
Mayhew [35]	1998	Denmark (UK–Danish collaboration)	Cohort (placental stereology)	Diabetes	Placentas from normal, high-altitude and diabetic pregnancies	Lowland normal placentas	Tier 4	Intervascular layer thickness
Mayhew [37]	1993	Denmark (UK–Danish collaboration)	Cohort (placental stereology)	Diabetes (White A-F/R)	89 placentas (55 diabetic; 34 controls)	Control deliveries	Tier 4	Placental oxygen diffusive conductance
Olofsson [90]	1986	Sweden	Prospective matched cohort (planned vaginal delivery)	PGDM; GDM	92 (46 diabetic; 46 matched controls)	Non-diabetic women matched for age, parity, GA, type of labour, scalp sampling and epidural	Tier 2 (intrapartum CTG reported separately)	Intrapartum FHR (ominous patterns); fetal scalp pH; cord pH
Shannon [91]	1986	USA	Prospective cohort	Diabetes (mostly PGDM)	31 (18 IDM; 13 controls)	Non-diabetic	Tier 4	Cord EPO; erythroid progenitors
Singer [92]	1986	USA	Retrospective autopsy comparison	PGDM; GDM	54 autopsies (16 IDM; 38 non-diabetic)	Autopsied fetuses and infants of non-diabetic mothers	Tier 4	Blinded morphometric hepatic erythropoiesis index
Widness [93]	1981	USA	Prospective cohort	Diabetes (PGDM/GDM)	89 (61 IDM; 28 controls)	Non-asphyxiated AGA infants	Tier 4	Umbilical plasma EPO
Fadel [94]	1979	USA	Cross-sectional (amniocentesis)	Diabetes	119 (19 diabetic; 59 controls)	Healthy third-trimester pregnancies	Tier 4	Amniotic fluid acid-base
Datta [95]	1977	USA	Prospective comparative cohort (elective caesarean)	PGDM (White B–R); GDM (n = 2)	60 (30 diabetic; 30 healthy)	Healthy non-diabetic women, elective CS	Tier 2	Umbilical arterial and venous pH, pO2, pCO2, base excess

Abbreviations: BE, base excess; CPR, cerebroplacental ratio; EPO, erythropoietin; GDM, gestational diabetes mellitus; HIE, hypoxic-ischaemic encephalopathy; MCA, middle cerebral artery; Mod-MPI, modified myocardial performance index; NE, neonatal encephalopathy; NRBC/nRBC, nucleated red blood cells; STV, short-term variation; T1D/T2D, type 1/type 2 diabetes; UA, umbilical artery; UV, umbilical vein. Studies are grouped by diabetes population and ordered by year of publication. Sample sizes are as reported; for registry studies, the number of births or deliveries is given.

3.4. Tier 2: Umbilical Acid–Base Status and Oxygenation at Birth

Thirty studies compared umbilical acid–base status, or umbilical oxygenation, between pregnancies with and without diabetes (Table 3). These outcomes describe acute compromise at birth and do not, by themselves, indicate hypoxic-ischaemic brain injury. Pregestational diabetes. In nine of 10 studies with data for pregestational diabetes, acid–base status at birth was less favourable than in controls, statistically significantly in eight. Fetal hypoxia defined by cord pH, lactate, or base deficit occurred in 20.0% of pregnancies with pregestational diabetes versus 4.7% of controls [71]; type 1 diabetes was associated with umbilical arterial pH < 7.10 (adjusted OR 3.35, 95% CI 1.79–6.27) [77]; and pregestational diabetes was associated with cord pH < 7.1 in a registry of 1.7 million deliveries (crude OR 1.69, 1.34–2.13) [76]. Mean pH was 0.03–0.05 units lower in type 1 diabetes in two further studies [39,65] and 0.03 units lower, with higher pCO2, in a third [64]; median pH was 0.06 units lower at elective caesarean section in a fourth [67]; cord pH and base excess were lower in pregestational diabetes in another [62]; and pH was 0.04 units lower, without statistical significance, in one cohort [70]. One study at elective caesarean section under epidural anaesthesia found slightly higher pH and base excess with diabetes [68]. GDM. Of 17 GDM comparisons, 13 showed a less favourable value in GDM, statistically significantly in 9; two showed a more favourable value [39,46], and two identical values or no numerical data [41,77]. Where differences were found, they were small: mean or median pH was 0.02–0.03 units lower, including at elective caesarean section without labour [26,44,47]; base deficit was modestly greater [26,29]; umbilical venous oxygen saturation was lower and lactate higher in well-controlled GDM delivered by elective caesarean section [25] and in obese women with GDM [28]; and a registry found a small excess of cord pH < 7.1 (crude OR 1.11, 1.03–1.20) [76]. In the largest intrapartum cohort, umbilical arterial pH < 7.10 occurred in 5.3% of GDM pregnancies versus 2.6% of those with a normal glucose tolerance test, and base excess < −12 mmol/L in 2.9% versus 0.5% [23]; the composite outcome reported by the authors, which included a 5 min Apgar score < 7, was not used because it combines Tier 2 and Tier 3 criteria. In a matched cohort with tight metabolic control, pH < 7.20 was almost equally frequent (10% vs. 9.4%) [60]. Mixed or unspecified diabetes. Seven of eight studies found less favourable acid–base status, with pH lower by 0.01–0.08 units where reported, statistically significantly in four [81,84,88,95]; the largest difference occurred after spinal anaesthesia with maternal hypotension at elective caesarean section [95]. In a study of 17,358 validated paired cord samples, acidaemia was slightly more frequent with diabetes (22 of 574, 3.8%, versus 496 of 16,784, 3.0%, calculated from the published tables) [30], and pH was lower without statistical significance in two further studies [74,79]; one study found identical fetal scalp pH [90]. Exploratory pooling. Three GDM studies reported umbilical arterial pH as mean and standard deviation against a non-diabetic comparator [26,47,70]; in one, diet- and insulin-treated GDM groups were combined. The random-effects mean difference was −0.025 (95% CI −0.037 to −0.014 with the DerSimonian–Laird method and −0.041 to −0.010 with the Hartung–Knapp method; I² = 0%), with a 95% prediction interval of −0.10 to 0.05. Two further GDM studies reported medians and interquartile ranges [22,44,46]; adding them after conversion gave −0.014 (Hartung–Knapp 95% CI −0.050 to 0.022; I² = 75%). Neither estimate approaches a clinically relevant threshold of acidaemia.

Table 3. Tier 2 outcomes: umbilical acid–base status and oxygenation at birth.

Study	Diabetes Type	Comparator	Main Findings	Direction
Gestational diabetes only (n = 12)
Kayaalp [44]	GDM	Non-GDM	Cord arterial pH median 7.34 vs. 7.36 (p = 0.001); base excess not different	▲
Szablewska [46]	GDM	Non-GDM	pH higher in GDM (7.32 vs. 7.27, p = 0.020), pCO2 lower; pO2, BE, HCO3 not different	▼
Golshahi [26]	GDM	Non-GDM, elective CS	pH lower (7.33 vs. 7.36, p = 0.006) and base deficit greater in GDM; pCO2, pO2, HCO3 not different	▲
Gul Kara [47]	GDM	Normotensive non-GDM	Cord pH 7.33 vs. 7.35 (p = 0.039); cord PlGF and sFlt-1 lower in GDM	▲
Zhang [41]	GDM (±PE)	Uncomplicated pregnancies	Cord blood gas disturbances most marked in GDM+PE; data for 20 of 42 infants only, in supplement	◄►
Boskabadi [48]	GDM	Preterm infants of healthy mothers	Venous pH 7.25 vs. 7.28 (not significant)	▲
Bhorat [31]	GDM (poorly controlled)	Matched controls	Cord pH < 7.2 in 7% of GDM vs. 0% of controls	▲
Bianchi [28]	GDM (overweight/obese)	Normal-weight and non-GDM women	Obese women with GDM had lower UA and UV pO2 and UV O2 saturation/content than normal-weight controls	▲
Tarvonen [23]	GDM	Normal OGTT; no OGTT	UA pH < 7.10: 5.3% vs. 2.6%; BE < −12 mmol/L: 2.9% vs. 0.5% (published composite including 5-min Apgar < 7 not used)	▲
Beneventi [29]	GDM	Non-GDM uncomplicated vaginal births	Base deficit greater in GDM (OR 1.51, 1.04–2.18 for higher quartile)	▲
Taricco [25]	GDM	Normal pregnancies	UV O2 saturation 53.8 vs. 63.2% and O2 content lower, lactate higher; pH not different	▲
Drexel [60]	GDM	Matched non-diabetic pregnancies	UA pH < 7.20: 10% vs. 9.4%; pH < 7.10: none in either group	▲
Pregestational diabetes only (n = 5)
Agaoglu [62]	PGDM (T1D, T2D)	Low-risk pregnancies	Cord arterial pH and base excess lower in pregestational diabetes	▲
Verhaeghe [64]	PGDM (T1D, T2D)	Newborns of non-diabetic mothers	UA pH 7.26 vs. 7.29 (p = 0.003); pCO2 higher; BE not different	▲
Verner [65]	T1D	Non-diabetic	Cord pH 7.25 vs. 7.30 (p < 0.01)	▲
Ruozi-Berretta [67]	PGDM	Normal pregnancies	Umbilical arterial pH median 7.22 vs. 7.28 (p = 0.003); pCO2 higher in PGDM	▲
Ramanathan [68]	T1D	Healthy euglycaemic women (group 3)	UA pH 7.32 vs. 7.30; BE −1.0 vs. −3.0 mmol/L; no significant difference	▼
Gestational and pregestational, or type not specified (n = 13)
Seyhanli [70]	PGDM; GDM (diet/insulin)	Non-diabetic controls	Cord pH 7.33 (pregestational), 7.33 (diet GDM), 7.35 (insulin GDM) vs. 7.37 (controls); p = 0.66	Pregestational ▲; GDM ▲
Kopteeva [71]	PGDM; GDM	Non-diabetic controls	Fetal hypoxia (cord pH < 7.10, lactate > 6.1 mmol/L or BE < −9 mmol/L): pregestational 20.0%, GDM 7.5%, controls 4.7% (p = 0.008)	Pregestational ▲; GDM ▲
Mahon [74]	Diabetes (PGDM/GDM)	Non-diabetic AGA newborns	UA pH 7.22 vs. 7.26 (not significant); pH < 7.20 in 24% vs. 16%; pO2 similar	▲
Beyerlein [76]	PGDM; GDM	No diabetes	2008–2016: PGDM crude OR 1.69 (1.34–2.13); GDM OR 1.11 (1.03–1.20)	Pregestational ▲; GDM ▲
Castelijn [77]	T1D; T2D; GDM	Non-diabetic controls	T1D: UA pH < 7.20 aOR 1.88 (1.23–2.87), <7.10 aOR 3.35 (1.79–6.27); T2D and GDM not significantly increased for low pH	Pregestational ▲; GDM ◄► (mean pH 7.24 vs. 7.24)
Zaigham [30]	PGDM; GDM	Non-diabetic	Acidaemia 22/574 (3.8%) vs. 496/16,784 (3.0%), calculated from published tables (not tested); cord lactate not different	▲
Stanescu [79]	Diabetes (type NS)	Uncomplicated pregnancies	Cord pH 7.32 vs. 7.33 and BE −4.92 vs. −5.04 mmol/L (not significant)	▲
Reif [81]	GDM; T1D	Non-diabetic with non-reassuring FHR	Arterial pH 7.215 vs. 7.250 (p = 0.007); scalp pH ≤ 7.25 14.1% vs. 7.1% (NS); Apgar similar	Diabetes ▲
Namavar Jahromi [84]	Insulin-requiring diabetes	Healthy pregnancies	UA pH 7.22 vs. 7.24 (p = 0.004); pCO2 higher; base excess not different	▲
Loukovaara [39]	T1D; GDM (insulin)	Normal pregnancies	UA pH 7.23 (T1D) and 7.27 (GDM) vs. 7.26 in controls; pCO2 higher in T1D; pO2 similar	Pregestational ▲; GDM ▼
Tan [88]	Diabetes (PGDM/GDM)	Matched non-diabetic	UA pH 7.241 vs. 7.283 (p < 0.05); acidosis (pH < 7.2) not more frequent	▲
Datta [95]	PGDM (White B–R); GDM (n = 2)	Healthy non-diabetic women, elective CS	UA pH 7.20 vs. 7.28 and BE −5.67 vs. −1.40 mEq/L after spinal anaesthesia; no difference after general anaesthesia	Diabetes ▲
Olofsson [90]	PGDM; GDM	Non-diabetic women matched for age, parity, GA, type of labour, scalp sampling and epidural	Mean scalp pH 7.34 vs. 7.34; cord acidosis 7/26 vs. 0/16 (subset only)	Diabetes ◄►

Abbreviations: BE, base excess; CPR, cerebroplacental ratio; EPO, erythropoietin; GDM, gestational diabetes mellitus; HIE, hypoxic-ischaemic encephalopathy; MCA, middle cerebral artery; Mod-MPI, modified myocardial performance index; NE, neonatal encephalopathy; NRBC/nRBC, nucleated red blood cells; OR, odds ratio; STV, short-term variation; T1D/T2D, type 1/type 2 diabetes; UA, umbilical artery; UV, umbilical vein. Direction: ▲ less favourable in the diabetic group; ▼ more favourable; ◄► no clear difference. Studies reporting GDM and pregestational diabetes separately show the direction for each population.

3.5. Tier 1: Diagnosed or Coded Neonatal Encephalopathy, HIE, or Seizures

Twelve studies compared a Tier 1 outcome between pregnancies with and without diabetes (Table 4). Definitions ranged from coded HIE or neonatal seizures in registries and birth certificates to neonatal encephalopathy treated with therapeutic hypothermia [24] and brain injury on term-equivalent magnetic resonance imaging [27]; no study reported encephalopathy staging, so these outcomes are diagnosed or coded rather than confirmed hypoxic-ischaemic injury. Pregestational diabetes. All five studies with sufficient events found an increased risk, statistically significant in each: HIE adjusted OR 5.09 (2.19–11.80) in a UK cohort of 53,649 pregnancies [4]; HIE or neonatal seizures adjusted OR 3.41 (2.58–4.49) for type 1 and 2.54 (1.13–5.69) for type 2 diabetes in 1.34 million Swedish births, after adjustment for maternal body mass index [63]; neonatal encephalopathy adjusted RR 2.42 (2.02–2.91), and 2.11 (1.49–2.98) for cases treated with hypothermia, in approximately 3 million Californian births [24]; neonatal seizures within 72 h adjusted OR 4.32 (1.62–11.51) in a population-based case-control study of Colorado birth certificates, adjusted only for vaginal delivery with herpes infection, post-term delivery, and fetal distress [86]; and neonatal seizures adjusted OR 1.51 (1.15–1.98) in a propensity-matched analysis of US birth certificates [72]. A matched type 1 diabetes cohort found HIE in 1.9% versus 0.6% of controls (OR 3.0), but with only four events, this comparison was classified as uninformative [61]. In the Swedish cohort, the association in type 1 diabetes persisted after restriction to term, normally grown infants of normotensive mothers [63]. GDM. Point estimates were in an adverse direction in six of seven informative comparisons, but most were small and imprecise: HIE 0.3% versus 0.2% [4]; neonatal seizures adjusted OR 1.06 (0.92–1.21) and unadjusted OR 1.53 (0.88–2.65) in two birth-certificate analyses [72,86]; and HIE 9.2% versus 7.8% in a preterm intensive care population with unspecified diagnostic criteria [50]. Statistically significant associations came only from the Californian cohort (adjusted RR 1.24, 1.14–1.34; 1.25, 1.07–1.46 for cases treated with hypothermia), about half that for pre-existing diabetes [24], and from a study restricted to small-for-gestational-age infants (unadjusted RR 5.61, 1.25–25.18, for clinically defined HIE) [53]. Among very preterm infants, brain injury on magnetic resonance imaging was slightly less frequent with GDM (intraventricular haemorrhage 10.6% vs. 13.8%) [27]. Two GDM comparisons were uninformative: no HIE in either group [58] and two cases of neonatal encephalopathy in 5150 births [23]. Unspecified diabetes. In a New Jersey registry, diabetes alone was associated with a non-significant increase in neonatal seizures (adjusted OR 1.37, 0.92–2.05), whereas diabetes with coexisting hypertension was significantly associated (2.30, 1.31–4.04) [82].

Table 4. Tier 1 outcomes: diagnosed or coded neonatal encephalopathy, hypoxic-ischaemic encephalopathy, or neonatal seizures.

Study	Diabetes Type	Comparator	Main Findings	Direction
Gestational diabetes only (n = 5)
Sappler [27]	GDM	GA/sex/scanner-matched preterm infants	No difference in any type of brain injury (intraventricular haemorrhage 10.6% vs. 13.8%); FA higher (enhanced maturation) in GDM; neurodevelopment at 24 months equal	▼
Zhang [50]	GDM	Non-GDM preterm	HIE 9.2% vs. 7.8%, not significant overall or in any GA subgroup	▲
Chen [53]	GDM	SGA infants of non-GDM mothers	HIE RR 5.61 (1.25–25.18) in GDM vs. non-GDM SGA infants (unadjusted)	▲
Tarvonen [23]	GDM	Normal OGTT; no OGTT	Neonatal encephalopathy: 2 cases in 5150 births	(◄►)
Sarkar [58]	GDM A1 (diet)	Non-diabetic	No HIE in either group	(◄►)
Pregestational diabetes only (n = 2)
Rizvi [61]	T1D	Matched healthy controls	HIE 1.9% vs. 0.6%, OR 3.0, NS (4 events); “hypoxia” 9.3% vs. 3.7%, p = 0.049	(◄►)
Cnattingius [63]	T1D; T2D	No diabetes	HIE/seizures OR 3.41 (2.58–4.49) T1D and 2.54 (1.13–5.69) T2D, adjusted incl. BMI	Pregestational ▲
Gestational and pregestational, or type not specified (n = 5)
Karkia [4]	PGDM; GDM	No diabetes	HIE PGDM 1.2% vs. 0.2%, aOR 5.09 (2.19–11.80); GDM 0.3% vs. 0.2%, not significant	Pregestational ▲; GDM ▲
Liang [72]	PGDM; GDM	Neonates without seizures	PGDM aOR 1.51 (1.15–1.98); GDM aOR 1.06 (0.92–1.21)	Pregestational ▲; GDM ▲
Bandoli [24]	Pre-existing diabetes; GDM	No diabetes	Pre-existing diabetes aRR 2.42 (NE) and 2.11 (NE+hypothermia); GDM aRR 1.24 (1.14–1.34) and 1.25 (1.07–1.46)	Pregestational ▲; GDM ▲
Potti [82]	Diabetes (type NS)	No diabetes or hypertension	Diabetes alone: seizures aOR 1.37 (0.92–2.05); diabetes plus hypertension aOR 2.30 (1.31–4.04)	Diabetes ▲
Hall [86]	PGDM; GDM	Term neonates without seizures	Neonatal seizures (≤72 h): pre-existing diabetes aOR 4.32 (1.62–11.51), adjusted for vaginal delivery with herpes, post-term delivery and fetal distress; GDM unadjusted OR 1.53 (0.88–2.65)	Pregestational ▲; GDM ▲

Abbreviations: BE, base excess; CPR, cerebroplacental ratio; EPO, erythropoietin; GDM, gestational diabetes mellitus; HIE, hypoxic-ischaemic encephalopathy; MCA, middle cerebral artery; Mod-MPI, modified myocardial performance index; NE, neonatal encephalopathy; NRBC/nRBC, nucleated red blood cells; OR, odds ratio; RR, relative risk; STV, short-term variation; T1D/T2D, type 1/type 2 diabetes; UA, umbilical artery; UV, umbilical vein. Direction (point estimate, regardless of statistical significance): ▲ higher risk in the diabetic group; ▼ lower risk; ◄► no clear difference; (◄►) fewer than five events in total, classified as uninformative and not counted. For Hall et al., the estimate from the final multivariable model is shown; the abstract of that report gives OR 4.30 (1.64–11.27). Estimates are adjusted where the study reported an adjusted estimate.

3.6. Tier 4: Chronic and Antepartum Markers of Fetal Hypoxic Stress

Forty-four studies compared a chronic or antepartum marker between pregnancies with and without diabetes (Table 5). After overlapping reports were counted once and two studies with hypertensive comparators were set aside, 33 of 37 independent sources showed an adverse direction in the diabetic group (statistically significant in 26), three no difference, and one changes interpreted as compensatory adaptation. Erythropoietin, nucleated red cells, polycythaemia, and erythropoiesis (15 studies; 15 adverse in direction, 11 statistically significant). In type 1 diabetes, umbilical venous erythropoietin was about 12-fold higher than in controls (294 vs. 23.8 mU/mL) and correlated with maternal HbA1c [40]; amniotic fluid erythropoietin was higher than in controls (median 14.0 vs. 6.3 mU/mL) and correlated with umbilical arterial pH and pO2 [38]; cord erythropoietin exceeded the control range in 36% of infants of diabetic mothers [93]; and cord erythropoietin and erythrocyte zinc protoporphyrin were higher in fetuses of insulin-treated mothers [87]. In three smaller studies, cord erythropoietin was higher with diabetes but not significantly so [65,83,91], and two found significantly higher cord erythropoietin in GDM [54,57]. Nucleated red cells were increased in infants of diabetic mothers in two studies [84,89], in poorly controlled GDM [31] and in two studies of GDM without intrapartum compromise [57,59] and were higher but not significantly so in preterm infants of mothers with GDM [48]. Polycythaemia (venous haematocrit ≥ 65%) occurred in 29.4% of infants of mothers with pregestational diabetes versus 5.9% of pair-matched controls [69], and hepatic erythropoiesis at autopsy was excessive in large infants of diabetic mothers [92]. Placental histology and morphometry (11 studies; 6sixof seven independent sources adverse). Changes interpreted as hypoxic were villous immaturity and placental nucleated red cells in GDM [56], altered villous morphology in well-controlled GDM [25], chorangiosis, villous immaturity and ischaemic changes in GDM [51,52,57], chorangiosis and, after exclusion of hypertensive disorders, chronic hypoxic patterns in a large pathology series [78], and reduced villous membrane diffusing capacity in type 1 diabetes [66]. Three stereology studies from one collaboration described increased oxygen diffusive conductance and thinner fetal plasma layers, which their authors interpreted as compensatory adaptation; they were counted as one source and coded as adaptation rather than as adverse [35,36,37]. Two studies comparing GDM with non-diabetic hypertensive pregnancies found similar vascular lesions; they were not counted, because they do not estimate the effect of diabetes [49,55]. Doppler velocimetry (six studies; four adverse in direction, three statistically significant). Cerebroplacental or cerebroplacental–uterine ratios were lower [44,62], or venous flow was redistributed [71], in three studies; in a fourth, the cerebroplacental ratio was slightly lower in GDM but slightly higher in pregestational diabetes [70]; two found no difference [43,88]. Antepartum fetal heart rate and cardiac function (seven studies; three of four independent sources adverse). The myocardial performance index was impaired in four studies from one group, counted as one source [31,32,33,34]; antepartum computerised cardiotocography showed reduced variability in GDM [42] and pregestational diabetes [67]; and fetal electrocardiographic heart time intervals did not differ in well-controlled GDM [45]. Intrapartum fetal heart rate findings (ZigZag patterns and late decelerations more frequent in GDM [23]; ominous patterns in 17.4% versus 10.9% in a matched cohort [90]) were not classified as Tier 4 and are considered with the evidence for Hypothesis A (Section 3.7). Stillbirth autopsy (four studies; four adverse in direction, three statistically significant). Among stillbirths, fetal cardiomyopathy or cardiomegaly [73,85] and thymic stress involution [80] were more frequent with diabetes, and placental disease was a similarly frequent cause of death (27% vs. 25%) [75]; in these studies, hypoxia was inferred rather than measured. Amniotic fluid pH was slightly lower with diabetes (7.04 vs. 7.07, not significant) [94]. Two studies used comparators that may exaggerate differences, namely autopsied infants who died of other conditions [92] and fetuses selected as non-hypoxic [87]; excluding them left 31 of 35 independent sources adverse. No Tier 4 study also reported a Tier 1 outcome in the same infants.

Table 5. Tier 4 outcomes: chronic and antepartum markers of fetal hypoxic stress.

Study	Diabetes Type	Comparator	Main Findings	Direction
Erythropoietin, nucleated red cells, polycythaemia, and erythropoiesis (n = 15)
Boskabadi [48]	GDM	Preterm infants of healthy mothers	NRBC/100 WBC 121.7 vs. 45.5 (p = 0.156)	▲
Mohd Idris [54]	GDM	Normal pregnancies	Cord EPO higher in GDM; BFU-E erythroid progenitors increased (94.2 vs. 73.6, p = 0.010)	▲
Madazli [57]	GDM	OGTT-negative women matched for GA, age and BMI	Cord EPO 48.6 vs. 31.8 mU/mL; NRBC 7.0 vs. 1.3/100 WBC; villous immaturity 59% vs. 9%, chorangiosis 27% vs. 5% (blinded histology)	▲
Yeruchimovich [59]	GDM	AGA infants of non-diabetic women	Absolute NRBC median 0.56 (LGA) and 0.13 (AGA) vs. 0.0005 × 109/L; diabetes an independent predictor	▲
Verner [65]	T1D	Non-diabetic	Cord EPO 45.7 vs. 41.9 IU/L (p = 0.058)	▲
Teramo [38]	T1D	Healthy controls at CS	AF EPO 14.0 vs. 6.3 mU/mL (p < 0.0001); 13.5% >63 mU/mL; EPO correlated with UA pH (r = −0.49) and pO2 (r = −0.62)	▲
Widness [40]	T1D	Non-diabetic	UV EPO 294 vs. 23.8 mU/mL (p < 0.001); correlated with maternal HbA1c (r = 0.57)	▲
Mimouni [69]	PGDM (White B–R)	Infants of non-diabetic mothers matched for GA, mode of delivery, Apgar group, sampling site/time and cord clamping	Polycythaemia (venous Hct ≥ 65%) 29.4% vs. 5.9% (p < 0.03); NRBC 13 vs. 3/100 WBC (p < 0.002)	▲
Cetin [83]	PGDM; GDM	Normoglycaemic controls	Cord EPO 12.7 vs. 12.3 mIU/mL (not significant); polycythaemia related to maternal β-hydroxybutyrate rather than EPO	▲
Namavar Jahromi [84]	Insulin-requiring diabetes	Healthy pregnancies	NRBC 560 vs. 202/µL (p < 0.001)	▲
Lott [87]	Insulin-treated (type NS)	Non-hypoxic control fetuses	Cord EPO and erythrocyte ZnPP/heme higher, ferritin lower in fetuses of insulin-treated mothers	Diabetes ▲
Vatansever [89]	T1D; GDM	Healthy term infants	Absolute NRBC higher in IDM; serum EPO higher but not significant	▲
Shannon [91]	Diabetes (mostly PGDM)	Non-diabetic	Mean cord EPO 56.3 vs. 44.4 mU/mL (not significant); EPO above range in 1/18 vs. 2/13	▲
Singer [92]	PGDM; GDM	Autopsied fetuses and infants of non-diabetic mothers	Hepatic erythropoiesis normal in normally grown IDM, excessive in large IDM at term (blinded morphometry)	Diabetes ▲
Widness [93]	Diabetes (PGDM/GDM)	Non-asphyxiated AGA infants	EPO above control range in 22/61 IDM (36%); EPO correlated with insulin (r = 0.73)	▲
Placental histology and morphometry (n = 11)
Mor [49]	GDM (±PE)	Isolated pre-eclampsia	Placental vascular lesion profile in GDM+PE similar to or milder than isolated PE; composite neonatal outcome lower with GDM	◄► ‡
Dasgupta [51]	GDM	Screen-negative pregnancies	Placental NRBC 28.5% vs. 0%; chorangiosis 81% vs. 24%; syncytial knots 86% vs. 19% (blinded)	▲
Sutreja [52]	GDM	Uncomplicated pregnancies	Chorangiosis 58% vs. 0%; syncytial knots 82% vs. 2%; fibrinoid necrosis 62% vs. 4%	▲
Kovo [55]	GDM (±hypertensive disorders)	Non-GDM hypertensive pregnancies	GDM+H vs. non-GDM+H: similar placental vascular lesions and neonatal outcome	◄► ‡
Taricco [25]	GDM	Normal pregnancies	Altered villous morphology on placental histology	▲
Daskalakis [56]	GDM	Consecutive normal pregnancies	Villous immaturity and fetal NRBC significantly more frequent in GDM placentas	▲
Jauniaux [66]	T1D	Uncomplicated elective CS	Specific diffusing capacity reduced (1.32 vs. 1.72, p = 0.032); adaptive changes in macrosomic subgroup	▲
Mayhew [36]	Diabetes (White class D)	Uncomplicated pregnancies	Diabetic placentas heavier, shorter fetal plasma distances, larger capillaries	◇
Stanek [78]	Maternal diabetes (PGDM/GDM)	Remaining placentas	Chorangiosis more common in diabetes; preuterine chronic hypoxic pattern significant only after excluding hypertensive conditions	▲
Mayhew [35]	Diabetes	Lowland normal placentas	Fetal plasma distance reduced and layers more irregular in diabetes, interpreted as adaptation to hypoxia	◇
Mayhew [37]	Diabetes (White A-F/R)	Control deliveries	Total diffusive conductance 7–25% higher in diabetic placentas (compensatory adaptation)	◇
Doppler velocimetry (n = 6)
Emiralioglu [43]	GDM	Non-GDM	All indices including CPR not different (p = 0.99); no brain-sparing redistribution	◄►
Kayaalp [44]	GDM	Non-GDM	CPR and cerebroplacental–uterine ratio lower, UA-PI higher in GDM	▲
Agaoglu [62]	PGDM (T1D, T2D)	Low-risk pregnancies	CPR and cerebroplacental–uterine ratio lower in pregestational diabetes	▲
Seyhanli [70]	PGDM; GDM (diet/insulin)	Non-diabetic controls	Doppler indices similar; CPR 2.38 (PGDM) and 2.13–2.18 (GDM) vs. 2.26 in controls (not significant)	Pregestational ▼; GDM ▲
Kopteeva [71]	PGDM; GDM	Non-diabetic controls	Redistribution of umbilical venous flow to the fetal liver and reduced ductus venosus shunt fraction in pregestational diabetes	▲
Tan [88]	Diabetes (PGDM/GDM)	Matched non-diabetic	Abnormal UA resistance index not more frequent	◄►
Antepartum fetal heart rate and cardiac function (n = 7)
Danciu [42]	GDM	Low-risk pregnancies	STV lower and LTV < 5 bpm more frequent in GDM than controls (p = 0.0077); pathological decelerations rare and not different	▲
Kömmel [45]	GDM	Healthy controls	Heart time intervals not different; HRV parameters differed but did not predict adverse outcome; HbA1c only predictor	◄►
Bhorat [31]	GDM (poorly controlled)	Matched controls	nRBC > 10/100 WBC in 19% vs. 0%; impaired Mod-MPI correlated with haemoglobin and haematocrit	▲
Bhorat [32]	GDM (well-controlled)	Matched normal pregnancies	MPI higher in GDM (p < 0.0001); cord pH < 7.15 in 7% of GDM, none in controls	▲
Bhorat [33]	Gestational impaired glucose tolerance (diet)	Matched normal pregnancies	MPI higher in GIGT (p < 0.0001); MPI predicted adverse outcome (AUC 0.96)	▲
Bhorat [34]	GDM (poorly controlled, insulin)	Matched normal pregnancies	Mod-MPI 0.59 vs. 0.38 (p < 0.0001); E/A lower; MPI > 0.52 predicted adverse outcome	▲
Ruozi-Berretta [67]	PGDM	Normal pregnancies	Antepartum cCTG: accelerations, high variation and STV lower	▲
Stillbirth autopsy (n = 4)
Lynch [73]	PGDM; GDM	Non-diabetic stillbirths	Fetal cardiomyopathy 46.7% vs. 3.3% (RR 14.0, 3.23–60.65); interpreted by authors as pathway to chronic hypoxia	▲
Page [75]	PGDM; GDM	Non-diabetic stillbirths	Placental disease as cause 27% (diabetic) vs. 25% (non-diabetic), NS; maternal medical complications more often causal	▲
Edwards [80]	Diabetes (PGDM/GDM)	Matched non-diabetic stillbirths	Thymic starry-sky pattern 11/20 vs. 4/22 (p = 0.03); smaller placentas; consistent with subacute stress	▲
Russell [85]	T1D; GDM	Unexplained non-diabetic stillbirths	Cardiomegaly 26% vs. 22% (LGA) and 0% (AGA) non-diabetic stillbirths; heavier hearts, lighter brains	Diabetes ▲
Amniotic fluid acid–base status (n = 1)
Fadel [94]	Diabetes	Healthy third-trimester pregnancies	Amniotic fluid pH 7.04 vs. 7.07 (not significant)	▲

Abbreviations: BE, base excess; CPR, cerebroplacental ratio; EPO, erythropoietin; GDM, gestational diabetes mellitus; HIE, hypoxic-ischaemic encephalopathy; MCA, middle cerebral artery; Mod-MPI, modified myocardial performance index; NE, neonatal encephalopathy; NRBC/nRBC, nucleated red blood cells; RR, relative risk; STV, short-term variation; T1D/T2D, type 1/type 2 diabetes; UA, umbilical artery; UV, umbilical vein. Direction (point estimate, regardless of statistical significance): ▲ adverse direction in the diabetic group; ▼ favourable direction; ◄► identical values or no numerical data; ◇ changes interpreted by the authors as compensatory adaptation (cluster counted once); ‡ comparator with another pregnancy complication, not counted. Studies are grouped by marker; studies reporting more than one Tier 4 marker are listed under their main marker.

3.7. Certainty of Evidence and Direction of Effect

Table 6 and Figure 2 summarise the certainty of evidence and Table 7 the direction of effect for each study; domain-level reasons for each rating are given in Supplementary Table S5. Certainty was very low for an association between pregestational diabetes and Tier 1 outcomes. The evidence is observational and was downgraded for indirectness, because the outcomes were coded or clinical diagnoses that do not distinguish hypoxic-ischaemic from other encephalopathy or seizures; although four of the five informative adjusted estimates exceeded 2, certainty was not rated up, because it had been rated down. Risk of bias was judged not serious: all five informative studies were population-based or large cohort or case-control studies, and none was rated high overall concern, although adjustment ranged from extensive maternal and obstetric covariates in four to three variables (vaginal delivery with herpes infection, post-term delivery, and fetal distress) in one [86]. Certainty was very low for an association between GDM and Tier 1 outcomes, very low for clinically relevant acidaemia in GDM, and low for less favourable acid–base status in pregestational diabetes. For chronic markers, certainty was very low for every marker, including erythropoietin in type 1 diabetes. Of the 74 studies, 19 informed only Hypothesis A (acute compromise at birth), 33 only Hypothesis B (chronic or antepartum markers), 11 both, and 11 reported only a Tier 1 outcome (Supplementary Table S4). Evidence for Hypothesis A came almost entirely from umbilical cord gases; only three studies also reported intrapartum fetal heart rate monitoring or fetal scalp sampling [23,81,90]. No study measured a chronic marker, an acute intrapartum event, and a Tier 1 outcome in the same pregnancies, so the combined chronic-plus-acute model (Hypothesis B) could not be assessed. In sensitivity syntheses, (a) restricting to the 52 studies rated low or moderate concern left the direction of every tier unchanged (Tier 1: seven of eight informative sources adverse in direction, with statistically significant estimates confined to pregestational diabetes, the Californian registry, and the study restricted to small-for-gestational-age infants; Tier 2: 23 of 26; Tier 4: 21 of 22); (b) counting overlapping reports separately did not change the direction of any marker; (c) restricting GDM acid–base evidence to elective caesarean or uncomplicated deliveries [25,26,29] retained small adverse differences in all three studies; and (d) excluding the two Tier 4 studies with atypical comparators left 31 of 35 sources adverse.

Table 6. Certainty of evidence (GRADE) by outcome tier, population and chronic marker.

Outcome	Population	Studies (Participants with Diabetes)	Direction of Effect	Risk of Bias	Inconsistency	Indirectness	Imprecision	Other	Certainty
Tier 1: HIE, NE or neonatal seizures (diagnosed or coded)	Pregestational	6 [4,24,61,63,72,86] (>15,000)	▲ in 5/5 informative, all statistically significant (1 uninformative, 4 events); adjusted OR/RR 1.5–5.1	Not serious (population-based or large studies; adjustment limited in one)	Not serious	Serious (coded or clinical diagnoses)	Not serious	None (large effect not used after rating down)	Very low ⊕◯◯◯
Tier 1: HIE, NE or neonatal seizures (diagnosed or coded)	GDM	9 [4,23,24,27,50,53,58,72,86] (>300,000)	▲ in 6/7 informative, statistically significant in 2 (registry aRR 1.24; SGA-only study); ▼ in 1; 2 uninformative	Serious (restricted populations; unadjusted estimates)	Not serious	Serious (coded diagnoses; restricted populations)	Serious (small, imprecise estimates; few events)	—	Very low ⊕◯◯◯
Tier 2: acid–base status at birth	Pregestational	10 [39,62,64,65,67,68,70,71,76,77] (>10,000)	▲ in 9/10 (statistically significant in 8)	Not serious	Not serious	Not serious	Not serious	—	Low ⊕⊕◯◯
Tier 2: clinically relevant acidaemia	GDM	17 (>38,000)	▲ in 13/17 (statistically significant in 9; small differences); pooled pH MD −0.025 (HK 95% CI −0.041 to −0.010; PI −0.10 to 0.05)	Serious	Serious	Serious (small differences of uncertain clinical meaning)	Not serious	—	Very low ⊕◯◯◯
Tier 4: erythropoietin	Type 1/pregestational or insulin-treated	6 [38,40,65,87,91,93]	▲ in 6/6 (statistically significant in 4); large differences in type 1 diabetes	Serious	Not serious	Not serious	Not serious	None (large effect not used after rating down)	Very low ⊕◯◯◯
Tier 4: erythropoietin and nucleated red cells	GDM	6 [31,48,54,57,59,83]	▲ in 6/6 (statistically significant in 4); small studies	Serious	Not serious	Not serious	Serious	—	Very low ⊕◯◯◯
Tier 4: nucleated red cells, polycythaemia, and erythropoiesis	Pregestational or mixed	4 [69,84,89,92]	▲ in 4/4	Serious	Not serious	Serious	Serious (small, older studies)	—	Very low ⊕◯◯◯
Tier 4: placental histology and morphometry	All	11	▲ in 6/7 independent sources; stereology cluster coded as adaptation; 2 studies with hypertensive comparators not counted	Serious	Serious	Serious	Serious	—	Very low ⊕◯◯◯
Tier 4: Doppler velocimetry	All	6	▲ in 4/6 (statistically significant in 3)	Serious	Serious	Serious	Serious	—	Very low ⊕◯◯◯
Tier 4: antepartum fetal heart rate and cardiac function	All	7 (4 from one group, counted once)	▲ in 3/4 independent sources	Serious	Not serious	Serious	Serious	Single-group evidence	Very low ⊕◯◯◯
Tier 4: stillbirth autopsy findings	All	4	▲ in 4/4 (statistically significant in 3; hypoxia inferred)	Serious	Serious	Serious	Serious	—	Very low ⊕◯◯◯
Chronic marker, acute event and Tier 1 injury in the same fetus	All	0	Not studied	—	—	—	—	—	No evidence

Abbreviations: aRR, adjusted relative risk; GDM, gestational diabetes mellitus; GRADE, Grading of Recommendations Assessment, Development and Evaluation; HK, Hartung–Knapp; HIE, hypoxic-ischaemic encephalopathy; MD, mean difference; NE, neonatal encephalopathy; OR, odds ratio; PI, prediction interval; RR, relative risk; SGA, small for gestational age. Observational evidence started at low certainty. Direction of effect: ▲, point estimate in the adverse direction in the diabetic group (higher risk of the outcome, less favourable acid–base status or a higher or more abnormal chronic marker); ▼, point estimate in the favourable direction in the diabetic group. Direction refers to the point estimate, regardless of statistical significance. “▲ in x/y” indicates that x of y informative studies, or independent evidence sources where overlapping reports were counted once, showed an adverse direction; the number with a statistically significant result is given in parentheses. Certainty was not rated up for a large effect when it had been rated down for any domain.

Figure 2. Evidence-synthesis map of hypoxic-ischaemic outcomes in pregestational and gestational diabetes, showing the direction of effect and GRADE certainty for each outcome tier and population.

Table 7. Effect-direction plot by outcome tier and diabetes population.

Study	Population	Tier 1	Tier 2	Tier 4	Overall Concern
Gestational diabetes only (n = 30)
Danciu [42]	GDM	–	–	▲	High
Emiralioglu [43]	GDM	–	–	◄►	High
Kayaalp [44]	GDM	–	▲	▲	Moderate
Kömmel [45]	GDM	–	–	◄►	High
Szablewska [46]	GDM	–	▼	–	Moderate
Golshahi [26]	GDM	–	▲	–	Low
Gul Kara [47]	GDM	–	▲	–	Moderate
Zhang [41]	GDM	–	◄►	–	High
Boskabadi [48]	GDM	–	▲	▲	High
Mor [49]	GDM	–	–	◄► ‡	Moderate
Sappler [27]	GDM	▼	–	–	Low
Zhang [50]	GDM	▲	–	–	High
Dasgupta [51]	GDM	–	–	▲	Moderate
Sutreja [52]	GDM	–	–	▲	High
Bhorat [31]	GDM	–	▲	▲	Moderate
Bianchi [28]	GDM	–	▲	–	Low
Chen [53]	GDM	▲	–	–	Moderate
Mohd Idris [54]	GDM	–	–	▲	Moderate
Tarvonen [23]	GDM	(◄►)	▲	–	Low
Bhorat [32]	GDM	–	–	▲	High
Bhorat [33]	GDM	–	–	▲	High
Kovo [55]	GDM	–	–	◄► ‡	Moderate
Beneventi [29]	GDM	–	▲	–	Low
Bhorat [34]	GDM	–	–	▲	High
Taricco [25]	GDM	–	▲	▲	Low
Daskalakis [56]	GDM	–	–	▲	Moderate
Madazli [57]	GDM	–	–	▲	Moderate
Sarkar [58]	GDM	(◄►)	–	–	Moderate
Yeruchimovich [59]	GDM	–	–	▲	Moderate
Drexel [60]	GDM	–	▲	–	Moderate
Pregestational diabetes only (n = 12)
Rizvi [61]	Pregestational	(◄►)	–	–	Moderate
Agaoglu [62]	Pregestational	–	▲	▲	Moderate
Cnattingius [63]	Pregestational	Pregestational ▲	–	–	Moderate
Verhaeghe [64]	Pregestational	–	▲	–	Moderate
Verner [65]	Pregestational	–	▲	▲	Moderate
Jauniaux [66]	Pregestational	–	–	▲	High
Ruozi-Berretta [67]	Pregestational	–	▲	▲	High
Teramo [38]	Pregestational	–	–	▲	Moderate
Mayhew [36]	Pregestational	–	–	◇	High
Ramanathan [68]	Pregestational	–	▼	–	Moderate
Widness [40]	Pregestational	–	–	▲	Moderate
Mimouni [69]	Pregestational	–	–	▲	Moderate
Gestational and pregestational, or type not specified (n = 32)
Seyhanli [70]	Mixed/unspecified	–	Pregestational ▲; GDM ▲	Pregestational ▼; GDM ▲	High
Kopteeva [71]	Mixed/unspecified	–	Pregestational ▲; GDM ▲	▲	Moderate
Karkia [4]	Mixed/unspecified	Pregestational ▲; GDM ▲	–	–	Moderate
Liang [72]	Mixed/unspecified	Pregestational ▲; GDM ▲	–	–	Moderate
Bandoli [24]	Mixed/unspecified	Pregestational ▲; GDM ▲	–	–	Low
Lynch [73]	Mixed/unspecified	–	–	▲	High
Mahon [74]	Mixed/unspecified	–	▲	–	Moderate
Page [75]	Mixed/unspecified	–	–	▲	High
Beyerlein [76]	Mixed/unspecified	–	Pregestational ▲; GDM ▲	–	Moderate
Castelijn [77]	Mixed/unspecified	–	Pregestational ▲; GDM ◄►	–	Moderate
Zaigham [30]	Mixed/unspecified	–	▲	–	Low
Stanek [78]	Mixed/unspecified	–	–	▲	Moderate
Stanescu [79]	Mixed/unspecified	–	▲	–	Moderate
Edwards [80]	Mixed/unspecified	–	–	▲	High
Reif [81]	Mixed/unspecified	–	Diabetes ▲	–	Moderate
Potti [82]	Mixed/unspecified	Diabetes ▲	–	–	Moderate
Cetin [83]	Mixed/unspecified	–	–	▲	Moderate
Namavar Jahromi [84]	Mixed/unspecified	–	▲	▲	Moderate
Russell [85]	Mixed/unspecified	–	–	Diabetes ▲	High
Hall [86]	Mixed/unspecified	Pregestational ▲; GDM ▲	–	–	Moderate
Lott [87]	Mixed/unspecified	–	–	Diabetes ▲	Moderate
Loukovaara [39]	Mixed/unspecified	–	Pregestational ▲; GDM ▼	–	Moderate
Tan [88]	Mixed/unspecified	–	▲	◄►	Moderate
Vatansever [89]	Mixed/unspecified	–	–	▲	Moderate
Mayhew [35]	Mixed/unspecified	–	–	◇	High
Mayhew [37]	Mixed/unspecified	–	–	◇	High
Olofsson [90]	Mixed/unspecified	–	Diabetes ◄►	–	Moderate
Shannon [91]	Mixed/unspecified	–	–	▲	Moderate
Singer [92]	Mixed/unspecified	–	–	Diabetes ▲	High
Widness [93]	Mixed/unspecified	–	–	▲	Moderate
Fadel [94]	Mixed/unspecified	–	–	▲	High
Datta [95]	Mixed/unspecified	–	Diabetes ▲	–	Moderate

▲ adverse direction in the diabetic group; ▼ favourable direction; ◄► identical values or no numerical data; (◄►) fewer than five events, uninformative; ◇ compensatory adaptation; ‡ comparator with another pregnancy complication, not counted; –, not assessed. Tier 1, diagnosed or coded neonatal encephalopathy, HIE or seizures; Tier 2, umbilical acid–base status or oxygenation at birth; Tier 4, chronic or antepartum markers. Studies reporting gestational and pregestational diabetes separately show the direction for each population. Overall concern as defined in Section 2.5 and Table 1. Direction refers to the point estimate, regardless of statistical significance (Section 2.8). Intrapartum fetal heart rate findings are not shown as Tier 4 (Section 2.7).

4. Discussion

The principal finding of this review is that no included study documented a chronic hypoxic marker, a defined acute intrapartum event, and diagnosed or coded hypoxic-ischaemic brain injury in the same fetus. Every element of the proposed “chronic substrate plus acute trigger” model (Hypothesis B) has been observed separately, mostly in different populations and largely in pregestational rather than gestational diabetes, but the model as a whole has not been tested. Within that limit, the 74 included studies show a gradient between pregestational diabetes and GDM. In pregestational diabetes, diagnosed or coded hypoxic-ischaemic injury was associated with diabetes in all five informative studies, including three large population-based cohorts adjusted for maternal and obstetric factors and a population-based case-control study with limited adjustment [4,24,63,86], with effects mostly of about two- to five-fold; acid–base status at birth was less favourable in most studies; and the most consistent chronic marker, erythropoietin, was elevated mainly in type 1 diabetes and correlated with maternal glycaemic control and umbilical arterial pH [38,40]. In the Swedish cohort, the association in type 1 diabetes persisted after restriction to term, normally grown infants of normotensive mothers [63], which argues against explanation solely by fetal size or hypertensive disease. In GDM, by contrast, estimates were mostly in an adverse direction but small and imprecise, and statistically significant associations came only from a registry based on administrative codes, in which the association was about half that for pre-existing diabetes and residual confounding by fetal size [96] or maternal obesity [63] cannot be excluded [24], and from an unadjusted comparison restricted to small-for-gestational-age infants [53]. Acid–base differences in GDM were small, about 0.02–0.03 pH units, and of uncertain clinical meaning. They were observed even at elective caesarean section without labour [25,26], which suggests a modest antepartum shift rather than intrapartum compromise, and in the largest validated cord-gas dataset, which combined pregestational diabetes and GDM, acidaemia was only slightly more frequent with diabetes (3.8% vs. 3.0%) [30]. Composite “asphyxia” and Apgar-based outcomes cannot serve as proxies for brain injury: in one type 1 diabetes cohort, 26.7% of infants met a composite asphyxia definition, yet none developed HIE [97].

The evidence bearing on the two hypotheses is also unbalanced. With Tier 3 outcomes excluded, Hypothesis A (an acute intrapartum event in a previously uncompromised fetus) rests almost entirely on umbilical cord gases. Cord gases describe the acid–base state at the moment of birth but cannot show when acidaemia developed: a low pH at delivery may follow an acute intrapartum event, a chronic antepartum process with superimposed labour, or an antepartum shift without labour, as suggested by the small differences found at elective caesarean section [25,26]. Only three studies also reported intrapartum fetal heart rate monitoring or fetal scalp sampling, and none linked an intrapartum event to a Tier 1 outcome in the same infants. The available data are therefore compatible with Hypothesis A but cannot distinguish it from Hypothesis B, and the small acid–base differences in GDM, observed even without labour, are at least as consistent with a modest antepartum shift as with intrapartum compromise.

Chronic markers are consistent with, but do not establish, a chronic hypoxic component. Erythropoietin, which is not stored and is synthesised in response to hypoxaemia, offers the most direct physiological evidence. Elevated levels were reported in type 1 or insulin-treated diabetes by different groups over four decades [38,40,87,93], although three smaller studies found smaller, non-significant differences [65,83,91], and GDM-specific data came from two small studies [54,57]. Polycythaemia and nucleated red cells, which also reflect increased erythropoiesis, were more frequent in infants of diabetic mothers [59,69], although polycythaemia did not correlate with maternal HbA1c in the only matched study [69]. Hytinantti et al. [98] proposed that the combination of erythropoietin and leptin might help estimate the duration of fetal hypoxia; this proposal rests on a single study without a concurrent control group and should be regarded as a hypothesis for future research, not as a tool for timing injury. Placental findings were frequent but heterogeneous. Some studies described features read as hypoxic injury, such as villous immaturity, chorangiosis, and nucleated red cells [25,51,52,56,57,78], whereas stereological studies described increased diffusive conductance read as adaptation [35,36,37]; the same tissue changes can therefore be interpreted either as a cause of fetal hypoxia or as compensation for it, and no study linked them to acid–base status or neurological outcome in the same infants. Doppler and fetal cardiac indices are indirect: impaired myocardial performance was reported by only one research group [31,32,33,34], and Doppler redistribution was inconsistent. Fetal hyperinsulinaemia raising oxygen consumption independently of placental perfusion remains a plausible complementary mechanism [31,93]. Plausibility, however, is not evidence of causation, and certainty was very low for every chronic marker (Table 6). Glycaemic control may be a more relevant modifier than the diagnosis itself. Erythropoietin correlated with maternal HbA1c in type 1 diabetes [38,40], cord-gas evidence of fetal hypoxia was concentrated in pregestational rather than gestational diabetes within the same cohort [71], and in a single small study of poorly controlled GDM, nucleated red cells were raised alongside fetal cardiac dysfunction [31]. Whether poorly controlled GDM approaches the hypoxic burden of pregestational diabetes remains an open question, because no study compared the two on the same outcome. Glycaemic control may also not be fully protective: in well-controlled GDM delivered by elective caesarean section, umbilical venous oxygen saturation was lower and lactate higher than in controls [25]. Quality assessment showed that the Newcastle–Ottawa Scale rewards the method of outcome ascertainment rather than the validity of the outcome definition: when the adequacy of each hypoxia-related outcome definition was judged separately, overall concern was low in only eight of 74 studies. Several studies excluded fetal growth restriction or pre-eclampsia by design, which improves internal validity but removes the population in which chronic hypoxic injury is most plausible. Independence was also limited: four fetal cardiac studies came from one group and three placental studies from one collaboration, and each cluster was counted once in the primary synthesis. Three studies from the Helsinki programme, whose GDM criteria (at least two abnormal OGTT values) were stricter than IADPSG criteria [38,39,40], enrolled women in different periods and were counted separately.

5. Strengths and Limitations

This review combined diagnosed or coded neonatal brain injury, acid–base status at birth, and chronic markers within one synthesis, classified outcomes by clinical meaning, required a concurrent non-diabetic comparator, and judged outcome-definition adequacy separately from standard risk-of-bias tools. Its limitations are substantial. First, the eligibility criteria, outcome tiers, appraisal approach, and synthesis methods were defined after the first analysis; these post hoc choices were applied to all records but may have been influenced by knowledge of the included studies. Second, the definitions of HIE and neonatal seizures were heterogeneous and frequently administrative, and no study reported encephalopathy staging. Third, 25 full-text reports, mostly older or regional publications, could not be retrieved, and several concerned erythropoietin, nucleated red cells, or placental findings. Because the retrieved studies of these markers were almost uniformly adverse in direction, the missing reports are unlikely to reverse the direction of the Tier 4 findings, but they could change their precision and the proportion of statistically significant results. In addition, the search strategies were developed by one author and checked against the PRESS checklist by a co-author who is not an information specialist, so peer review of the search was not independent of the review team. Then, most biomarker, Doppler, and placental studies were small, unadjusted, and from high-income settings, and many predate current diagnostic criteria for GDM. Two Tier 4 studies used comparators that may exaggerate differences (autopsied infants who died of other conditions and fetuses selected as non-hypoxic), although excluding them did not change the direction of the findings. Additionally, type 2 diabetes was reported separately in only three studies. Moreover, heterogeneity precluded a full meta-analysis; vote counting by direction of effect does not weigh study size or precision and counts small, non-significant differences in the same way as large ones, which is why statistical significance is reported alongside direction, and the exploratory pooled estimate of umbilical pH rests on three studies, with a wide prediction interval. Finally, only English-language publications were included.

6. Conclusions

Pregestational diabetes, particularly type 1 diabetes, is associated with diagnosed or coded hypoxic-ischaemic encephalopathy, neonatal encephalopathy or seizures, less favourable acid–base status at birth, and markers of chronic fetal hypoxic stress; the certainty of this evidence is low for acid–base status at birth and very low for the other outcomes. In GDM, estimates for these outcomes mostly point in an adverse direction but are small and imprecise (very low certainty), and acid–base differences at birth are small. No study has shown a chronic process and an acute event combining to cause brain injury in the same fetus, so current evidence neither establishes nor excludes a chronic contribution in an individual case. Future prospective studies should measure chronic markers, intrapartum events, and neurological outcomes in the same GDM and pregestational cohorts, with standardised encephalopathy staging, neuroimaging, and follow-up beyond the neonatal period.