Manage gestational and pregestational diabetes mellitus in pregnancy

In one line

Diabetes in pregnancy is a continuum from gestational diabetes (GDM) to pre-existing type 1/type 2 disease, and management turns on separating the two early, driving glucose to tight targets, and managing the predictable obstetric harms — congenital anomaly, macrosomia, stillbirth, neonatal hypoglycaemia — rather than the glucose number alone.

This chapter assumes the diagnostic and first-line groundwork in diabetes & glucose-tolerance basics and the hypertension-in-pregnancy groundwork; it concentrates on subtype mechanism, the judgement calls, the named regimens and where guidelines disagree. Prevention of the pre-eclampsia that travels with diabetes is in pre-eclampsia-prevention-aspirin.

Aetiology and pathophysiology — why the subtypes behave differently

That pregnancy is diabetogenic is assumed (placental lactogen, cortisol and progesterone drive a roughly second-half insulin resistance). What matters is classifying the disease by mechanism and reading each subtype's distinct obstetric signature off that mechanism. Four phenotypes sit under "hyperglycaemia first detected in pregnancy" (HFDP) plus the two pregestational types, and they do not behave alike.

• Type 1 diabetes (autoimmune β-cell failure). Absolute insulinopenia, so the patient is ketosis-prone and the dominant non-obstetric emergency is DKA — which in pregnancy arrives faster, at lower glucose, and kills the fetus (see below). Glycaemic lability is the rule: the same placental anti-insulin hormones that raise requirements also magnify the swing, so these women have the widest glycaemic variability and the strongest case for CGM. The clinical consequence: anomaly risk tracks periconceptional HbA1c, and intrapartum/neonatal hypoglycaemia is most severe because tight third-trimester control is hardest to hold.
• Type 2 diabetes (insulin resistance + relative insulinopenia). Now the commonest pregestational type in the SA public sector, riding the obesity epidemic. The mechanism is insulin resistance, so these women are older, heavier, more often hypertensive and dyslipidaemic, and frequently already vasculopathic at booking despite a "milder"-sounding diagnosis. The trap is complacency: outcome data repeatedly show T2DM pregnancies do as badly or worse than T1DM for stillbirth and perinatal death, because the disease is under-recognised, often poorly controlled at conception, and managed less intensively. The mechanism drives the consequence: established vascular disease (nephropathy, retinopathy, ischaemic heart disease) is what drives superimposed pre-eclampsia, growth restriction and the stillbirth excess — not the glucose alone.
• Overt (pregestational) diabetes first detected in pregnancy. A booking fasting ≥7.0 mmol/L, 2-hour OGTT ≥11.1 mmol/L or HbA1c ≥6.5% is pre-existing disease unmasked, not GDM. Mechanistically it has been hyperglycaemic through organogenesis, so it carries the anomaly and retinopathy risk of pregestational disease and is managed as such — the single most important classification call to get right.
• Gestational diabetes — and its two mechanistic sub-phenotypes. GDM is not one thing. Physiologically it splits into a predominantly insulin-resistant phenotype (the obese, metabolic-syndrome woman — looks like early T2DM, the larger group) and a predominantly insulin-deficient/β-cell-dysfunction phenotype (leaner, impaired secretory reserve). This is not academic: the insulin-resistant subtype responds to metformin and lifestyle and carries the macrosomia/LGA and later-T2DM risk, whereas the insulin-deficient subtype reaches insulin sooner and behaves more like nascent autoimmune or MODY disease. Suspect MODY (especially glucokinase, GCK-MODY) in the lean woman with a strong autosomal-dominant family history and mild, stable fasting hyperglycaemia that barely moves on treatment — here aggressively insulinising an unaffected fetus can cause growth restriction, while an affected fetus is appropriately macrosomic; the management pivots on fetal genotype, inferred from serial growth.

Two mechanism→consequence links unify the obstetric harms. First, the timing of the insult selects the lesion. Periconceptional hyperglycaemia is teratogenic (neural tube, the near-pathognomonic caudal regression/sacral agenesis, and cardiac and renal anomalies) because it acts during organogenesis — which is exactly why GDM, arising after organogenesis, does not raise anomaly risk and overt diabetes does. Late hyperglycaemia instead drives fetal hyperinsulinaemia (the Pedersen hypothesis): maternal glucose crosses the placenta, the fetal pancreas hypertrophies, and fetal insulin acts as a growth factor → asymmetric macrosomia (shoulders and trunk, hence shoulder dystocia), organomegaly, polyhydramnios, delayed surfactant maturation (RDS at term), and the neonatal hypoglycaemia that follows when the maternal glucose supply is cut at delivery but the hyperplastic islets keep secreting. Second, the vasculopathic placenta does the opposite of macrosomia — in the woman with nephropathy or long-standing disease, micro-vascular placental insufficiency gives growth restriction, abnormal Dopplers and the stillbirth risk, so the same diagnosis can present as either a 4.5 kg baby or a growth-restricted one depending on the maternal vascular bed. Both possibilities have to be held in mind at once.

(The old White classification — lettered by age of onset, duration and end-organ disease — is no longer recommended by ACOG and you should not lead with it; the modern stratifier is presence and severity of vasculopathy, which predicts adverse outcome better than duration-based class.)

Assessment

The first discrimination is GDM versus overt (pregestational) diabetes first detected in pregnancy — a distinction the term "hyperglycaemia first detected in pregnancy" (HFDP) is meant to force. A fasting glucose ≥7.0 mmol/L, a 2-hour OGTT value ≥11.1 mmol/L or an HbA1c ≥6.5% at booking is overt diabetes, not GDM: it carries the anomaly and retinopathy risk of pre-existing disease and is managed as such.

• Who to screen, when. SA practice (SEMDSA, mirroring WHO 2013) is predominantly risk-factor-based selective screening rather than universal — a resource-driven divergence from IADPSG universal screening. High-risk women (prior GDM, prior macrosomia, obesity, first-degree family history, glycosuria, PMOS (previously PCOS) — see polycystic-metabolic-ovarian-syndrome) get a 75 g OGTT at booking; if normal, repeat at 24–28 weeks. The HAPO-derived IADPSG/WHO 2013 GDM thresholds are any one of: fasting ≥5.1, 1-hour ≥10.0, 2-hour ≥8.5 mmol/L. NICE NG3 deliberately uses a different two-point threshold (fasting ≥5.6 or 2-hour ≥7.8 mmol/L) — the two guideline sets disagree, and which net is used materially changes who is diagnosed. The diagnostic thresholds side by side — the same 75 g OGTT read against two different nets:

The lower IADPSG/WHO net (used in SA) diagnoses more women than NICE; the woman with fasting 5.3 and 2-hour 8.0 is GDM by SA/IADPSG but not by NICE — which is exactly the "overdiagnosis in a resource-limited system" debate below.

• Pre-existing diabetes booking work-up. Baseline HbA1c (also dates anomaly risk), renal function with albumin:creatinine ratio, dilated retinal screening, BP and a careful drug review — stop ACE-inhibitors/ARBs, statins and any GLP-1 receptor agonist or SGLT2-inhibitor preconceptionally.
• Interpretation. HbA1c quantifies anomaly risk: it climbs steeply above ~6.5% and a periconceptional HbA1c >86 mmol/mol (10%) is the NICE threshold at which pregnancy is actively advised against until improved. Proteinuria or rising creatinine flags nephropathy, which predicts superimposed pre-eclampsia and growth restriction.

The advanced assessment — atypical presentations and the judgement calls

Beyond the GDM-versus-overt split, the discriminations that matter are these:

• The threshold disagreement is a real management fork. A woman with fasting 5.3 and 2-hour 8.0 mmol/L is GDM by IADPSG/WHO and SA practice but not by NICE NG3. You must be able to say which net you are using and why — and recognise that the lower IADPSG net diagnoses more women, most of whom have the milder insulin-resistant phenotype that responds to lifestyle, which is the substance of the "overdiagnosis in a resource-limited system" debate.
• HbA1c misleads in exactly the women you most want to assess. The expanded red-cell mass and shortened erythrocyte survival of pregnancy lower HbA1c independently of glycaemia, and iron-deficiency anaemia falsely raises it while haemolysis, recent transfusion and late pregnancy falsely lower it — all common in the SA antenatal population. So HbA1c is a good booking anomaly-risk index but a poor monitoring tool in the third trimester; track capillary/CGM glucose, not serial HbA1c, near term. In a woman with a haemoglobinopathy (also common locally), HbA1c may be uninterpretable on standard assays altogether.
• Read the booking HbA1c as a teratogenicity dosimeter. It does not just "flag risk" — anomaly rate rises roughly continuously with periconceptional HbA1c, so a value of, say, 9–10% in a newly-presenting woman mandates a detailed anomaly scan (including fetal echocardiography at ~20–22 weeks and specific interrogation of the spine/sacrum for caudal regression), not routine surveillance. Conversely a near-normal booking HbA1c in known diabetes is genuinely reassuring about organogenesis and you can say so.
• The vasculopath hides in plain sight. A T2DM woman with microalbuminuria, a creatinine at the upper-normal limit, or background retinopathy is a different pregnancy from the metabolically-similar woman without end-organ disease: she is the one who develops early superimposed pre-eclampsia and FGR. Stratify by vasculopathy, look for it actively (ACR, creatinine, dilated fundoscopy, BP), and let it — not the glucose — drive the intensity of fetal surveillance and the aspirin decision.
• Retinopathy can accelerate because you tighten control. Rapid normalisation of long-standing hyperglycaemia, superimposed on pregnancy's own retinal effects, can transiently worsen proliferative retinopathy — which is why dilated screening is repeated (and why you do not crash the HbA1c in a woman who presents late with very poor control and untreated proliferative disease without ophthalmology input).
• Subtle/atypical pictures that change the plan: unexpectedly falling insulin requirements in the third trimester (placental insufficiency — investigate, see red flags); recurrent unexplained hypoglycaemia in early pregnancy (the physiological first-trimester dip, but also new Addison's or pituitary disease if profound); a lean woman with mild, treatment-resistant fasting hyperglycaemia and a dominant family history (think GCK-MODY before escalating insulin); and the vomiting T1DM woman with near-normal glucose who is actually in euglycaemic DKA.

Management

Management runs immediate → ongoing → long-term, with the obstetric and the glycaemic threads kept explicit.

Immediate (at diagnosis). Confirm GDM vs overt disease. Start glucose self-monitoring against capillary targets and a 1–2 week trial of diet and exercise for GDM. For pre-existing diabetes already pregnant, optimise insulin and add folic acid 5 mg daily to 12 weeks plus low-dose aspirin 150 mg nocte from 12 weeks (diabetes is a high-risk factor for pre-eclampsia — see pre-eclampsia-prevention-aspirin and the Intermediate hypertension groundwork).

Glycaemic targets (capillary plasma glucose; NICE NG3, broadly shared by ADA 2026):

Ongoing pharmacotherapy. For GDM, escalate stepwise: if diet fails after 1–2 weeks, metformin (titrated to 2–2.5 g/day); add or switch to insulin if targets are unmet. Start insulin first-line if the fasting glucose is ≥7.0 mmol/L at diagnosis or there is fetal macrosomia/polyhydramnios already. The guidelines genuinely diverge here: NICE and SMFM accept metformin first-line (cheap, oral, effective), whereas the ADA 2026 still names insulin the preferred agent because metformin and glibenclamide cross the placenta and insulin does not. In the SA public sector, metformin's oral route and cost make it pragmatically central; glibenclamide is largely abandoned given higher neonatal hypoglycaemia and macrosomia. Pre-existing T1DM/T2DM needs a basal-bolus regimen (e.g. isophane or detemir with a rapid analogue such as aspart), with insulin requirements typically falling in the first trimester then rising steeply to term.

Surveillance. Serial growth scans (macrosomia, polyhydramnios, and conversely growth restriction in vasculopaths), repeat retinal screening at 28 weeks (and 16–20 weeks if retinopathy present), and monthly HbA1c. Continuous glucose monitoring (CGM) is now standard-of-care in T1DM after CONCEPTT — see the controversy below.

Timing and mode of birth. Deliver pre-existing diabetes (T1/T2) at 37+0 to 38+6 weeks; deliver uncomplicated GDM no later than 40+6 weeks. Diabetes is not an indication for caesarean per se, but estimated fetal weight guides counselling about shoulder dystocia; a fraught instrumental or full-dilatation decision links to prolonged-second-stage and caesarean-at-full-dilatation. Run an intrapartum insulin–dextrose (sliding-scale) infusion to keep maternal glucose 4–7 mmol/L, which reduces neonatal hypoglycaemia.

Long-term / postnatal. Insulin requirements fall abruptly after delivery — halve pre-pregnancy doses in T1DM and usually stop all hypoglycaemics in GDM. Metformin and insulin are breastfeeding-compatible. GDM confers a high lifetime type 2 diabetes risk: arrange a fasting glucose or HbA1c at 6–13 weeks postpartum and annual screening thereafter, and counsel on contraception that does not delay this metabolic follow-up (see contraception-in-high-risk-women and postpartum contraception basics).

Subtype-specific management — what changes by phenotype

The generic ladder above is tailored by phenotype:

• Type 1 diabetes. Basal-bolus from the outset (or continue an insulin pump/closed-loop if she has one); plan for the U-shaped requirement curve — a first-trimester fall (with first-trimester nausea and the fetal/placental glucose drain making hypoglycaemia a real risk, so pre-empt it and prescribe glucagon) and a steep rise to roughly term, often to 2–3× booking doses. CGM is genuinely indicated here (below). She is the woman in whom euglycaemic DKA must be actively excluded whenever she vomits.
• Type 2 diabetes. The decision is not "metformin or insulin" but how aggressively to insulinise a woman who is already vasculopathic. MiTy is the trial that frames this: adding metformin to insulin improves glycaemia and cuts LGA but doubles SGA — so in a T2DM woman with a placenta already at risk of insufficiency, the SGA signal is a genuine reason to weigh metformin carefully and watch growth. Stop the non-insulin agents she may be on that are unsafe (SGLT2-inhibitors, GLP-1 RAs, sulfonylureas other than where unavoidable) and convert to insulin ± metformin.
• Overt diabetes detected in pregnancy. Manage as pregestational from the moment of diagnosis: anomaly scan + fetal echo, retinal screening, renal assessment, aspirin, and the earlier (37–38⁺⁶ week) delivery window — do not let the "first detected in pregnancy" label downgrade it to GDM surveillance.
• GDM, insulin-resistant phenotype. Lifestyle + metformin first; most are controlled without insulin. This is the group for whom the SA selective-screening, metformin-centred pathway is well-matched.
• GDM, insulin-deficient phenotype / suspected MODY. Reaches insulin sooner; in suspected GCK-MODY, anchor decisions on serial fetal growth as a proxy for fetal genotype rather than chasing the maternal number, because over-treating a mother carrying an unaffected (normoglycaemic) fetus causes iatrogenic FGR — a subtlety worth flagging rather than acting on blindly without specialist input.

Named insulin regimens and how they differ

"Start insulin" is incomplete without the named regimen and its titration. The three architectures are:

• Basal-bolus (multiple daily injections) — the default in pregnancy. A long/intermediate basal (isophane/NPH, or detemir — detemir is specifically supported in pregnancy by RCT data and has a flatter, more predictable profile than NPH with less nocturnal hypoglycaemia) plus a rapid-acting analogue at each meal (aspart or lispro — both placenta-non-crossing, licensed in pregnancy, and superior to soluble human insulin for post-prandial peaks and fewer hypos). Glargine is acceptable if a woman is already well-controlled on it, but detemir is the better-evidenced basal. Titrate against the specific number that is out of range: raise the basal for an elevated fasting glucose; raise the prandial dose for the elevated post-meal value at that meal.
• Continuous subcutaneous insulin infusion (CSII / pump), increasingly hybrid closed-loop. Smoother delivery and the substrate for automated insulin delivery; the AiDAPT trial supports closed-loop in T1DM pregnancy for time-in-range. Realistically restricted in the SA public sector by cost — name it as best-available rather than standard-of-care locally.
• Intrapartum insulin–dextrose infusion (sliding scale / VRIII). A separate regimen for labour: a variable-rate IV insulin infusion run alongside dextrose to hold maternal glucose 4–7 mmol/L, hourly capillary glucose, the goal being to prevent the reactive neonatal hypoglycaemia that tight intrapartum control averts. Well-controlled diet-only GDM in labour often needs only hourly monitoring, not the infusion.

The two situations where the regimen must be proactively rewritten:

• Antenatal corticosteroids. Betamethasone/dexamethasone for fetal lung maturity causes a predictable, transient maternal hyperglycaemia lasting roughly 72 hours, peaking within the first day. In an insulin-treated woman this commonly needs a substantial pre-emptive insulin increase — published cohorts find requirements rise on the order of ~80%, escalating over the first two to three days then tapering — delivered either as a stepped increase to her usual regimen or via a VRIII, with intensified monitoring. Never give steroids to an insulin-requiring diabetic without a glucose plan, and never let steroid-induced hyperglycaemia be misread as worsening disease.
• The U-shaped antenatal curve. Reduce doses in the first trimester to pre-empt hypoglycaemia; escalate steadily through the second and third; and then halve (T1DM) or stop (GDM) at delivery — the single commonest avoidable postnatal hypoglycaemic event is forgetting that the placenta, the source of the insulin resistance, has just left.

The SA/NDoH choice versus the international one

Where SA practice diverges, name the local choice: selective risk-based screening (not IADPSG universal), metformin-centred first-line oral therapy (cheap, oral, EML-listed) reserving insulin for failure/overt disease/early macrosomia, human insulins and isophane where analogues are restricted by formulary, point-of-care/NHLS glucose rather than universal CGM, and a district→regional→tertiary pathway in which a pregestational diabetic or a poorly-controlled GDM belongs at a unit with the relevant surveillance and neonatal back-up. The targets are international but the tools are rationed: manage to the best available locally, and document what the ideal would add.

Diabetic ketoacidosis in pregnancy — the metabolic emergency

DKA is an acute emergency and is easily mishandled. The pregnancy-specific facts:

• It develops faster and at lower glucose. Pregnancy is a state of accelerated starvation and relative insulin resistance with reduced buffering reserve, so ketoacidosis can establish within hours and at near-normal glucose — euglycaemic DKA accounts for a substantial minority of cases (up to roughly a third). A vomiting, tachypnoeic T1DM woman with a "normal" glucose can still be acidotic: check blood ketones and a venous gas, do not reach for an antiemetic alone. The diagnostic triad remains blood ketones ≥3.0 mmol/L (or significant ketonuria), bicarbonate <15 mmol/L and/or venous pH <7.3, with known or new diabetes.
• It is a fetal emergency. Maternal acidosis and volume depletion reduce uteroplacental perfusion; the CTG often shows a non-reassuring trace that frequently recovers as the mother is corrected — which is the crucial point: resuscitate the mother first, and resist an instinctive crash caesarean for an abnormal trace in an unstable, acidotic woman, because delivery into maternal acidosis worsens outcomes and the trace usually improves with maternal correction. Continuous fetal monitoring once viable, in parallel with maternal resuscitation.
• Management differs from the non-pregnant protocol in two ways. First, introduce dextrose early (alongside the fixed-rate insulin infusion) so that ketogenesis is switched off by insulin without inducing hypoglycaemia — essential in the euglycaemic case where glucose is already normal at presentation. Second, correct aggressively and monitor intensively: hourly capillary glucose and hourly blood ketones (aiming for a fall of at least ~0.5 mmol/L per hour), generous isotonic fluid resuscitation, careful potassium replacement as insulin drives it intracellularly, treat the precipitant (infection, missed insulin, hyperemesis, and — iatrogenically — β-sympathomimetic tocolysis and corticosteroids), and involve medicine/critical care and the neonatal team early.

The evidence & the controversy

Metformin's long-term safety is the live debate. It crosses the placenta, which is the basis of the ADA's caution. The MiTy RCT (type 2 diabetes, metformin vs placebo added to insulin) reduced large-for-gestational-age births (RR 0.65, 95% CI 0.43–0.99) and maternal weight gain but roughly doubled small-for-gestational-age births (13% vs 7%, RR 1.96, 95% CI 1.10–3.64) — the harm to weigh explicitly in T2DM. Its 24-month follow-up, MiTy Kids (Lancet Diabetes & Endocrinology, 2023), is broadly reassuring: metformin was not a predictor of BMI z-score at 24 months (mean difference −0.01, 95% CI −0.42 to 0.37, p=0.92). There is a sex-specific signal — male offspring had a significantly higher BMI trajectory between roughly 6 and 24 months (p=0.048). Short-term data are reassuring, but childhood-adiposity questions remain open and the longer MiTy Tykes (5–11 year) follow-up is awaited.

Treating early GDM. The TOBOGM trial (NEJM 2023) randomised women with GDM diagnosed before 20 weeks to immediate versus deferred treatment and found a modest reduction in a composite of adverse neonatal outcomes with immediate treatment — the first RCT evidence that early diagnosis and treatment helps, supporting booking OGTT in high-risk women (exactly SA's selective strategy).

CGM in type 1 diabetes. CONCEPTT (Lancet 2017) showed real-time CGM in pregnant T1DM women modestly improved time-in-range (68% vs 61%, ~1 extra hour/day in range) despite only a small HbA1c gain (−0.19%), yet still reduced large-for-gestational-age babies (OR 0.51), neonatal hypoglycaemia (OR 0.45) and NICU stay >24 h (OR 0.48) — about 6 women treated to prevent one NICU admission or one LGA baby. That a tiny HbA1c change drove real neonatal benefit shows that glycaemic variability, not just the mean, matters. In SA the constraint is cost and access: the benefit must be weighed against sensor affordability in a public system that cannot offer CGM universally.

The CONCEPTT data underpin the international consensus CGM targets in pregnancy (a different, more concrete output than an HbA1c): for pregnant women with type 1 diabetes, aim for time-in-range (3.5–7.8 mmol/L) >70%, time-below-range (<3.5 mmol/L) <4% (and <1% below 3.0), and time-above-range (>7.8 mmol/L) <25%. These are demanding numbers — even at term only about a third of women reach >70% TIR — but the dose–response is real: roughly every 5% gain in time-in-range tracks with better neonatal outcomes, which is the mechanistic reason CONCEPTT's small HbA1c change still moved hard endpoints. The role of CGM in T2DM/GDM is less well-evidenced and is not a justification for universal sensor funding in a constrained system. Automated insulin delivery (hybrid closed-loop) in T1DM pregnancy is supported by AiDAPT for time-in-range, but its place locally is limited by cost.

Glibenclamide retains a place only where insulin is genuinely unavailable; the weight of evidence (more neonatal hypoglycaemia, more macrosomia than metformin or insulin) has pushed it out of first-line use.

Landmark trials & key evidence

The studies that anchor a management plan, with their effect sizes. The recurring pattern: treating hyperglycaemia reliably reduces fetal-overgrowth morbidity (LGA, shoulder dystocia) but rarely moves a hard composite — so the benefit shows in the secondary outcomes.

A quick worked piece of the arithmetic. From ACHOIS, serious perinatal complications fell from 4% (control) to 1% (treatment): absolute risk reduction ARR = 4% − 1% = 3% = 0.03, so NNT = 1/ARR = 1/0.03 ≈ 34 — i.e. treat about 34 women with mild GDM to prevent one serious perinatal complication. From CONCEPTT, the reported number-needed-to-treat of about 6 to prevent one LGA baby or one NICU admission implies an absolute risk reduction of roughly 1/6 ≈ 0.17 (17 percentage points) on those endpoints — a large absolute effect for a sensor that barely moved the HbA1c, which is precisely why the trial reframed the target as glycaemic variability/time-in-range rather than the mean.

Worked viva — how to structure the answer

Consider a 34-year-old, BMI 38, known type 2 diabetes on metformin and glibenclamide, booking at 9 weeks, HbA1c 9.4%, BP 138/88, urine ACR raised. The management runs:

1. The clinical picture — this is poorly-controlled pregestational type 2 diabetes booking in the first trimester, with evidence of early vasculopathy (raised ACR); the HbA1c of 9.4% places her at materially increased risk of congenital anomaly, and the nephropathy predicts superimposed pre-eclampsia and growth restriction.
2. Immediate bundle — folic acid 5 mg, aspirin 150 mg nocte from 12 weeks, stop the glibenclamide and any ACE-I/ARB/statin/SGLT2-i/GLP-1, convert to insulin (basal-bolus, detemir + aspart) while continuing metformin only if her growth/SGA risk allows, optimise glucose to the capillary targets, dilated retinal screen and baseline renal assessment.
3. Anomaly and surveillance plan — detailed anomaly scan with fetal echocardiography at ~20–22 weeks and explicit interrogation of the spine/sacrum; serial growth scans expecting either macrosomia or (given the vasculopathy) FGR; repeat retinal screening.
4. Timing and intrapartum plan — deliver in the 37+0–38+6 window for pregestational disease (earlier if pre-eclampsia/FGR supervene), counsel on EFW and shoulder dystocia, run an intrapartum insulin–dextrose infusion to hold glucose 4–7 mmol/L.
5. The evidence — MiTy for the metformin-vs-SGA trade-off in T2DM, HAPO/ACHOIS for the continuum and the value of treatment (NNT ≈ 34), CONCEPTT/consensus targets for the glycaemic goal.
6. Anticipate the iatrogenic complications — neonatal hypoglycaemia (intrapartum control is the lever), euglycaemic DKA if she vomits, steroid-induced hyperglycaemia if preterm birth threatens, and the abrupt postnatal fall in requirements.
7. Close the loop — halve insulin at delivery, 6–13 week postnatal glucose/HbA1c, lifelong screening, contraception that does not delay follow-up, and pre-pregnancy optimisation counselling for next time.

Exam traps & red flags

• Calling overt diabetes "GDM." A booking fasting ≥7.0 mmol/L, OGTT 2-hour ≥11.1 mmol/L or HbA1c ≥6.5% is pre-existing diabetes — it needs anomaly scanning, retinal screening and aspirin, not reassurance.
• Forgetting the periconceptional bundle: 5 mg folate, HbA1c optimisation, and stopping teratogens (ACE-I/ARB, statins, GLP-1 RAs/SGLT2-i). Aspirin from 12 weeks is mandatory in a diabetic woman and is easily omitted.
• Insulin requirements that fall in late pregnancy are a red flag for placental insufficiency/impending fetal compromise, not improving control — investigate, don't celebrate.
• Trusting a third-trimester HbA1c. Pregnancy physiology and iron-deficiency anaemia distort HbA1c; monitor capillary/CGM glucose near term, not serial HbA1c.
• Under-treating type 2 diabetes because it "sounds milder" than type 1 — T2DM pregnancies do as badly or worse for stillbirth; the vasculopathy is the driver, not the label.
• The neonate is the emergency you cause: anticipate and screen for neonatal hypoglycaemia (and polycythaemia, jaundice, respiratory distress); tight intrapartum maternal control is the lever.
• Diabetic ketoacidosis in pregnancy can occur at near-normal glucose (euglycaemic DKA), develops faster, and is a fetal as well as maternal emergency — a sick vomiting T1DM woman needs ketones and a venous gas checked, not antiemetics alone, and the abnormal CTG usually corrects when you correct the mother.
• Crash-delivering the unstable DKA mother for an abnormal trace. Resuscitate the mother first; the trace typically recovers, and delivery into maternal acidosis worsens outcome.
• Giving steroids without a glucose plan. Antenatal corticosteroids cause ~72 h of hyperglycaemia needing a large pre-emptive insulin increase; do not misread it as worsening disease and do not omit the plan.
• Defaulting to caesarean for "diabetes": the indication is fetal size/obstetric, not the diagnosis; over-calling it is as wrong as missing a 4.5 kg estimated fetal weight.
• Over-insulinising suspected GCK-MODY — chasing the maternal number can cause iatrogenic FGR in a normoglycaemic fetus; let serial growth guide and seek specialist input.
• Losing the woman postpartum: no 6–13 week glucose test = a missed chance to catch type 2 diabetes; in SA this follow-up gap is a documented systemic failure. And forgetting to halve/stop insulin at delivery causes avoidable maternal hypoglycaemia.

Evidence anchors

• NICE NG3 — Diabetes in pregnancy: management from preconception to the postnatal period (recommendations)
• ADA Standards of Care in Diabetes 2026 — Section 15, Management of Diabetes in Pregnancy
• International Consensus on Time in Range — clinical CGM targets, including pregnancy (Diabetes Care 2019)
• TOBOGM — Treatment of Gestational Diabetes Mellitus Diagnosed Early in Pregnancy (NEJM 2023)
• MiTy Kids — 24-month follow-up of offspring exposed to metformin vs placebo (Lancet Diabetes Endocrinol 2023)
• CONCEPTT — Continuous glucose monitoring in type 1 diabetes pregnancy (Lancet 2017)
• Management of diabetic ketoacidosis in pregnancy (The Obstetrician & Gynaecologist, 2017)
• USPSTF — Low-dose aspirin to prevent pre-eclampsia (2021 recommendation)
• Hyperglycaemia first detected in pregnancy in South Africa — facts, gaps, opportunities (Frontiers 2022)
• SEMDSA / WHO 2013 GDM diagnostic criteria (endorsed in SA; no single stable canonical URL)
• National Department of Health, Guidelines for Maternity Care in South Africa (knowledgehub.health.gov.za) — SA selective-screening and treatment context