Prolactin in relation to gestational diabetes and metabolic risk in pregnancy and postpartum: A systematic review and meta-analysis

Context Pre-clinical evidence suggests that prolactin has important metabolic functions in pregnancy and postpartum, in addition to lactogenic actions. Objective To explore the relationship between prolactin and maternal metabolic outcomes in human pregnancy and postpartum, particularly in relation to gestational diabetes mellitus (GDM). Data sources MEDLINE via OVID, CINAHL plus, Embase. Study selection Eligible studies included women who were pregnant or up to 12 months postpartum, reporting at least one maternal serum prolactin level in relation to key metabolic outcomes including GDM, glycaemic parameters, obesity, and gestational weight gain. Data extraction Two independent reviewers extracted data. Data synthesis Twenty-six articles were included. Meta-analysis showed no relationship between maternal prolactin levels and GDM status, with a weighted mean difference of -2.14 ng/mL (95% CI -12.54 to 8.27 ng/mL, p=0.7) between GDM and controls in early pregnancy (n=3 studies) and -3.89 ng/mL (95% CI, -15.20 to 7.41 ng/mL, p=0.5) in late pregnancy (n=11 studies). In narrative synthesis of other outcomes (due to study heterogeneity and/or lack of data), prolactin levels were not associated with maternal glycaemic or weight-related parameters during pregnancy, but in the postpartum period (particularly with lactation) a high-prolactin environment was associated with low circulating insulin and beta-cell function, and increased insulin sensitivity. Conclusions Current evidence from human studies does not clearly support a relationship between prolactin and metabolic parameters during pregnancy, including with GDM status. Elevated prolactin was associated with lower insulin and beta-cell function and higher insulin sensitivity in the post-partum period, but the direction of causality remains unclear. Systematic review registration https://www.crd.york.ac.uk/prospero/, identifier [CRD42021262771].


Introduction
Human pregnancy is a period marked by profound reproductive and metabolic adaptations, including a progressive increase in maternal insulin resistance which is paralleled by increased maternal synthesis and secretion of insulin. Failure to sufficiently augment insulin secretion to overcome pregnancy-induced insulin resistance results in maternal gestational diabetes mellitus (GDM), defined as carbohydrate intolerance of variable severity with onset or first recognition during pregnancy (1). The resulting hyperglycaemia increases the risk for fetal macrosomia and obstetric complications (1,2). An improved understanding of the mechanisms that drive changes in maternal insulin resistance, and further insights into biomarkers which can facilitate early identification of women at risk of GDM, are needed to optimise prevention efforts and mitigate potential complications.
Prolactin (PRL) is a 199-amino acid polypeptide hormone produced by lactotrophs in the anterior pituitary gland. It signals through the PRL receptor, with signal transduction activating the Janus kinase-2 signal transducer and activator of transcription 5 (JAK-STAT5) pathway. Whilst the hormone is best known for its lactogenic effect on the female mammary gland, PRL receptors are also found in tissues important in metabolism, such as pancreatic beta-cells, hepatocytes, adipocytes, macrophages, and skeletal muscle (3). PRL alters insulin sensitivity, adipocyte function and lipid metabolism in vitro in both human and animal models (4). During gestation, rising levels of lactogenic hormones such as PRL and placentally-derived human placental lactogen (hPL) may contribute to systemic insulin resistance and reduced insulin binding, but have also been directly implicated in the parallel process of maternal pancreatic beta-cell proliferation and increased insulin secretion. Such findings initially emerged predominantly from animal work (5), but have been corroborated by results from in vitro human studies (6,7). As such, altered lactogen dynamics may contribute to the pathophysiology of insulin resistant conditions such as GDM (8,9). Postpartum, observational evidence consistently links breastfeeding to improved long-term maternal metabolic outcomes, such as reduced risk of progression to type 2 diabetes mellitus (T2DM). Such benefits may be mediated, at least in part, by lactation-induced changes to carbohydrate and lipid metabolism and adipocyte biology (3), with PRL as a central hormonal regulator (10).
Narrative reviews (which constitute the majority of the existing work in this area, and have produced many of the current mechanistic hypotheses) are often incomplete or reach subjective conclusions, and all rely heavily on preclinical research, often conducted in animal models. Multiple observational studies over several decades have explored various aspects of the relationship between PRL and gestational metabolic outcomes in pregnant human populations, but have not yet been effectively synthesised.
In this systematic review, we examine current evidence regarding the relationship between PRL and maternal metabolic outcomes in pregnancy and postpartum, particularly in relation to GDM and maternal glycaemia, as well as GDM risk factors. We provide mechanistic insights and examine the clinical implications of these findings.

Systematic review question
In pregnant women (participants) what is the relationship between PRL levels (exposures) and: (a) maternal gestational metabolic status/outcomes? (b) maternal metabolic outcomes up to 12 months postpartum?

Protocol and registration
A protocol for this review has been previously published (11). The review is part of a larger evidence synthesis examining lactogenic hormones in pregnancy and postpartum, was conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines, and is registered with the International Prospective Register of Systematic Reviews (PROSPERO), CRD42021262771.

Search strategy and databases
A systematic search strategy (Supplementary Material 1) combining MeSH terms and text words was developed using the OVID platform, in consultation with expert subject librarians, and was translated to other databases as appropriate. MEDLINE via OVID, MEDLINE ePub ahead of print, in-process, in-data review and other non-indexed citations via OVID, CINAHL plus, and Embase were searched from inception to 8 July 2021 (updated 9 May 2022).

Inclusion and exclusion Criteria
Selection criteria using a modified version of the Participant, Exposure, Comparison, Outcome and Study Type (PECOT) framework (11,12), established a priori, were used to determine the eligibility of articles for inclusion in this review. Using this framework, studies were included when the following criteria were fulfilled: participants were pregnant women and women up to 12 months postpartum, regardless of lactation status and with any comparison group (or no comparison); endogenous maternal serum PRL must have been measured and reported at least once during pregnancy and/or up to 12 months postpartum; and at least one of the key maternal outcomes below were reported: • GDM status during pregnancy, and diabetes status up to 12 months postpartum • Metabolic indices (continuous measurements) related to maternal glucose/lipid metabolism (e.g. glucose measurements on oral glucose tolerance test; insulin secretion; insulin sensitivity/resistance indices; beta-cell function) during pregnancy or postpartum • Body mass index/obesity, gestational weight gain • Postpartum weight change • Lipid profile There were no date limits for eligibility, but only articles with full text available in English were included. Eligible study types included cross-sectional, longitudinal cohort or case-control, and randomised controlled trials. Narrative and systematic reviews were excluded, but their bibliographies were examined to identify relevant articles.
Key exclusion criteria included: populations with pathological PRL elevation (e.g. prolactinoma) in pregnancy; studies involving exogenous administration of PRL; studies involving an intervention or procedure to manipulate PRL; studies involving medications known to affect PRL (e.g. dopamine agonists); studies in which PRL was only measured in another fluid (e.g. amniotic fluid or cord blood); studies focused on assisted reproductive technologies or primarily focused on women with other pregnancy pathologies (e.g. preeclampsia, placental dysfunction, stillbirth); as well as animal studies and in vitro/tissue culture studies. Commentaries, letters, conference abstracts, and case reports were also excluded.

Study selection and risk of bias assessment
Two independent reviewers (KR and RG) screened all articles on abstracts and full text and assessed methodological quality of included studies, with 10% of quality assessments performed in duplicate. Quality appraisal (risk of bias) was performed on Covidence using the Monash Centre for Health Research and Implementation (MCHRI) Evidence Synthesis Program critical appraisal tool (Supplementary Material 2), which is based on the Newcastle-Ottawa Scale for nonrandomised studies (13). Individual quality items were evaluated using a descriptive component approach to assess factors affecting external validity (methodology, inclusion/ exclusion criteria, and appropriateness of measured outcomes) and internal validity (attrition, detection, selection and reporting bias, confounding, statistical analyses, and study power). Studies that fulfilled all, most or few criteria were deemed to have low, moderate, and high risk of bias, respectively. Disagreements at any stage were resolved through discussion between reviewers to reach a consensus.

Data extraction
Data were manually extracted from all included studies by two independent reviewers using a specifically developed data extraction form in Microsoft Excel. Duplicate extraction was performed for 10% of studies, with no discrepancies identified. Information was collected on general details (authors, reference/ source, country, year of publication, study design, duration of follow-up), participants (baseline age, metabolic conditions, parity, body mass index [BMI], ethnicity, lactation status), PRL timepoints and values, PRL assay methodology, key maternal outcomes assessed in relation to PRL (unadjusted and adjusted, with consideration of covariates used), and conclusions.

Evidence synthesis and statistical analysis
Review Manager 5.4.1 software was used to perform metaanalysis for eligible outcomes. Where published papers contained insufficient data to be entered into meta-analysis, further details were sought from the authors. Random effects models were employed to generate weighted mean differences (WMD). Statistical heterogeneity was assessed using the I 2 test, with I 2 values of >50% indicating moderate to high heterogeneity. Sensitivity analyses were performed to explore the effects of studies with high risk of bias on the overall results. Recognising that older studies likely reflected a different clinical and research environment, sensitivity analysis was also performed with exclusion of studies published prior to 2000. For outcomes where meta-analysis was not possible (e.g. where studies were highly heterogeneous in methodology), narrative synthesis of results was performed. Data is presented in summary tables and in narrative format to describe the populations, exposures and key outcomes of the included studies. Forest plots and funnel plots have been used to present results from meta-analyses and publication bias assessments, respectively.

Search results
A total of 3922 results were retrieved from the initial database search. Following removal of duplicates, 2643 and 190 studies were excluded at abstract and full text screening, respectively, with reasons documented for excluded full texts ( Figure 1). Of note, the 51 studies excluded on the basis of English full text unavailability were disproportionately dated, with all but one published prior to 1997.
Of the 62 studies which met broader eligibility criteria for inclusion, 26 of these pertained to PRL in relation to maternal glycaemia (pregnancy/postpartum), GDM status, maternal weight or lipids, and were included in the present review. Meta-analysis was possible for exploring differences in early and late pregnancy PRL by GDM status, incorporating data from three and 11 studies for these timepoints, respectively (Figures 2A, B). All studies were observational in nature.

Risk of bias and publication bias assessments
Of the 26 studies included, five were deemed high risk of bias, 15 moderate, and six low (Tables 1-3). The main aspects contributing to high risk of bias were statistical analysis (inadequate detail, inadequate description of methodology, lack of adjustment for key confounding variables), and variability in outcome measurement and reporting; which were present in four and three of the five studies deemed high risk of bias, respectively. Visual inspection of funnel plots was not suggestive of publication bias for any of the analyses (Supplementary Material 3).

Prolactin in relation to GDM status and maternal glycaemic parameters in pregnancy
Fifteen studies (14-28) examined PRL in relation to GDM status (n=13 studies) and/or to maternal glycaemia in pregnancy

Prolactin in GDM vs controls in early pregnancy (≤24 weeks)
Four studies (15, 21, 23, 24) measured PRL in the first trimester of pregnancy and correlated this to the subsequent development of GDM. Three of these (15, 23, 24) contained sufficient data for meta-analysis ( Figure 2A), with pooled results suggesting no significant difference in early pregnancy PRL between GDM and non-GDM groups (WMD of -2.14 ng/mL, 95% CI -12.54 to 8.27 ng/mL, p=0.7) and moderate heterogeneity (I 2 = 49%, p het =0.1).
The fourth study (21) found significantly higher PRL in a multiracial cohort of GDM cases (n=104) compared with controls (n=213), representing the first large-scale prospective study of the association between early pregnancy PRL levels and GDM risk, and implicating PRL in the early pathophysiology of GDM. Results were reported in median and interquartile range format due to non-normal distributions for PRL, andwhen contactedthe authors were unable to provide the original data for inclusion in the meta-analysis.

Prolactin in GDM vs controls in late pregnancy (>24 weeks)
Thirteen studies compared PRL in late pregnancy between GDM and non-GDM pregnancies. Eleven of these contained sufficient data for meta-analysis, or this was supplied by the authors ( Figure 2B) (14-16, 18, 19, 23-28). Where more than one late pregnancy timepoint was available in a study, the latest was used. Pooled data from these 11 studies showed no significant difference in late pregnancy PRL between women with GDM and controls (WMD of -3.89 ng/mL, 95% CI -15.20 to 7.41 ng/mL, p=0.5), with moderate heterogeneity (I 2 = 64%, p het =0.001). Sensitivity analyses were performed stratifying by publication date (removing studies prior to vs after 2000) and risk of bias (removing studies deemed high risk of bias); with no major changes to overall effect or to heterogeneity.
A single study (24) examined the ratio of third trimester (29 week) to first trimester (12 week) PRL and compared this between GDM and non-GDM groups, finding no significant difference therein.

Prolactin in relation to maternal glycaemic parameters in pregnancy
Maternal glucose measurements or oral glucose tolerance test (OGTT) results were analysed in direct relation to PRL in three studies. Ekinci et al. (17) reported that 2 hour OGTT glucose values at 28 weeks were positively related to PRL at 35-39 weeks. In contrast, the remaining studies reported that PRL was not cross-sectionally associated with plasma glucose at 10-14 or 15-26 weeks (21) or with glucose area under the curve at 29 weeks (26).

FIGURE 2B
Forest plot showing meta-analysis of PRL levels in GDM vs non-GDM control women in late pregnancy (>24 weeks) -11 studies.  *Non-pregnant WHO diabetes criteria were used in this paper (despite the pregnant population), with "GDM" defined as fasting >7.8 or 2h >11.1 mmol/L, and "IGT" as fasting 6-7.8 or 2h 7.8-11.1 mmol/L. As such, the "GDM" group in this paper may have had overt diabetes, and the "IGT" group either GDM or overt diabetes; the "controls" may also have contained women with GDM according to appropriate pregnancy criteria.
Insulin and C-peptide were related to PRL in two studies. One found no relationship between fasting insulin and PRL at 29 weeks (26), while the other (21) reported that PRL correlated positively with maternal insulin and C-peptide at 10-14 weeks, but by 15-26 weeks, the relationship with insulin was attenuated and PRL was inversely related to C-peptide (likely due to differences in fasting status between the two timepoints (21)).
HbA1c was measured in relation to PRL in two studies (21,24), with one (21) reporting no relationship at 10-14 or 15-26 weeks, and the other (24) reporting an inverse association in both early (12 weeks) and late pregnancy (29 weeks).
Three studies reported the relationship between PRL and homeostatic model assessment of insulin resistance (HOMA-IR): one in early pregnancy, two in late pregnancy. All found no significant relationship (21,24,26).
Markers of beta-cell function (derived from insulin and glucose measurements) were reported in relation to PRL in three studies. One (24) reported a positive relationship between HOMA of beta-cell function (HOMA-b) and PRL in late pregnancy. In contrast, PRL was not associated with the insulinogenic index at 33-38 weeks (28), or with two similar derived measures of maternal beta-cell function at 29 weeks (26), in two other studies.
Only one group used gold-standard clamp techniques to directly measure insulin sensitivity (at 34-36 weeks) in a small cohort, finding no relationship with maternal PRL (20).

Prolactin in relation to maternal BMI and/or gestational weight gain in pregnancy
Five studies (Table 2) analysed PRL in relation to maternal pre-pregnancy BMI and/or gestational weight gain (GWG). Two reported lower PRL at 24-28 weeks in Korean women classified as overweight (BMI >23 kg/m 2 ) (25) and at 37 weeks in Chinese women classified as overweight/obese (BMI >24 kg/m 2 ) (31) compared with lean controls. For GWG, Lappas et al. (30) found no significant relationship, but described a trend toward lower PRL at delivery among non-obese, non-diabetic women in whom GWG exceeded recommended thresholds (compared with women with GWG within recommended ranges). The remaining two studies found no significant relationships between maternal PRL (measured at 16 and 27 weeks) and either pre-pregnancy BMI or GWG across a combined cohort of Chinese and Caucasian American women (32) or in a subset of the Caucasian American women only (29), after adjustment for multiple covariates.

Prolactin in relation to lipid profile parameters in pregnancy
Three studies analysed PRL in relation to maternal lipid profiles during pregnancy. Both Montelongo et al. (23) and Couch et al. (16) collected serial samples across pregnancy and found that (across all samples) PRL was significantly positively correlated with lipoprotein triglycerides (across all lipoprotein classes); likely reflecting the parallel tendency of both parameters to increase with advancing gestation.

Prolactin in relation to maternal glycaemia/metabolism during lactation and postpartum
Ten studies examined maternal serum PRL in relation to maternal metabolism in the postpartum period and/or during lactation (Table 3). These studies were particularly heterogeneous in their aims and methodology, precluding meta-analysis.

Postpartum prolactin in relation to maternal GDM status in pregnancy
Three studies examined postpartum PRL following pregnancies affected by GDM (at various timepoints between 2 weeks and 5 months postpartum, and in both lactating and nonlactating women). All three suggested no significant difference in maternal serum PRL according to GDM status (23, 28, 34).

Postpartum prolactin in relation to maternal BMI
Six studies examined maternal postpartum PRL in relation to BMI or overweight/obesity, of which three found negative associations and three found no relationship. Rasmussen et al. (38) showed that women with preconception overweight or obesity had a lower PRL response to infant suckling than their lean counterparts at 48 hours postpartum, and that maternal overweight/obesity preconception was an independent predictor of lower PRL response to suckling at 48 hours and 7 days postpartum. Similarly, Harreiter et al. (34) reported that prepregnancy BMI in lactating and non-lactating women at 3-5 months postpartum was negatively associated with PRL. Ren et al. (31) found lower PRL levels at both 37 weeks' gestation and 48 hours postpartum in women with overweight/obesity than lean controls; accompanied by significantly more delayed lactogenesis. In the three remaining studies, two reported that maternal BMI was not associated with serum PRL postpartum (regardless of lactation status (36), or in lactating women at 3-6 months postpartum (35)); the third found no relationship between BMI at delivery and either pre-feed PRL or PRL response to a feed at 4-5 days postpartum (33).  PRL pos assoc, sig.
Higher PRL levels assoc with lower pre-preg BMI and lower postpartum fasting insulin (univariate analyses, listed). After multivariate analysis, pre-hepatic betacell function and Stumvoll 1st phase insulin secretion index (but not BMI) independently and neg assoc with PRL levels.
Conclusion was that beta-cell function lower in lactating than non-lactating women (independent of BMI) and is inv assoc with PRL. Both lean and obese lactating mothers have lower IR. Authors suggest that good beta-cell plasticity (allowing beta-cell function to fall from high insulin production during preg to low production postpartum) may enable PRL to rise with permissive effect on lactation.  Both lactating and non-lactating women had low insulin. Fasted, lactating women had 2.6x higher basal EGP and 2.3x rates of lipolysis compared with nonlactating. When hyperinsulinaemic euglycaemic clamp applied (mimicking fed state), both groups suppressed lipolysis and EGP, but lactating women needed 36% less insulin to do so, suggesting postpartum insulin sensitivity may be further augmented by lactation. In lactating group, higher PRL was related to better insulin-mediated suppression of lipolysis, and lower intra-hepatic and circulating Tg. again at 48h postpartum preg (see Table 2) and 48h postpartum. At 48h postpartum overweight/obese 281.79 ± 87.61 vs. lean 392.96 ± 104.54, sig.

Rasmussen et al, 2014
significantly delayed onset of lactogenesis. Factors emerging as likely sig contributors to the lactogenesis delay (and poss also lower PRL levels) in the overweight/obese group were (a) higher late-preg leptin levels and (b) a slower fall in estrogen following delivery. PRL in late preg had been sig higher in those with NGT at 3mo postpartum than in those with postpartum pre-diabetes or DM. Higher late preg PRL independently predicted higher beta-cell function at 3 mo. PRL in late preg was independent predictor of risk of pre-DM or DM at 3mo PP (higher late preg PRL predicted lower dysglycaemia risk). Authors suggest that serum PRL in preg may provide novel insight into postpartum DM risk in young women, and suggest this might relate to known role of PRL in beta-cell mass expansion (here extending from preg into postpartum). NS difference in mean PRL between GDM and non-GDM women either in later preg or at 4-8 weeks postpartum, during lactation. No assoc found between PRL change (preg to postpartum) and change in AUC glucose (preg to postpartum). PRL did not change during OGTT (either in preg or postpartum).

Prolactin in relation to other aspects of maternal postpartum metabolism
Five studies examined PRL in relation to broader aspects of maternal postpartum metabolism, including continuous metabolic parameters.
Ozisik et al. (36) reported an inverse association between postpartum PRL and both HbA1c and C-peptide in a small cohort of women (n=22, 12 of whom were lactating). Harreiter et al. (34) studied 106 women (n=51 with a history of GDM, and 61 of whom were lactating) at 3-5 months postpartum. On univariate analysis, PRL at this time was negatively associated with maternal pre-pregnancy BMI, fasting glucose, fasting insulin and C-peptide, HOMA-IR, and beta-cell function. On multivariate regression, pre-hepatic beta-cell function and firstphase insulin secretion remained independently and negatively associated with PRL. Of note, when women were stratified according to lactation status, this inverse relationship was only seen in the lactating women. In contrast, Nurek et al. (35) found no relationship between basal PRL and fasting insulin in 37 lactating women at 3-6 months postpartum.
Only one study (37) used hyperinsulinaemic euglycaemic clamp techniques to relate maternal glucoregulatory physiology postpartum to PRL, comparing lactating and non-lactating women. Circulating insulin levels were low in both groups. Lactating women had higher rates of endogenous glucose production and lipolysis during fasting than the non-lactating group. When the clamp was supplied (mimicking the fed state), lactating women required 36% less insulin for suppression of lipolysis compared with non-lactating women. Of note, within the lactating group, higher PRL levels were associated with better insulin-mediated suppression of lipolysis (as well as lower intrahepatic triglyceride content and lower circulating triglycerides).
A single study (39) related pregnancy PRL measurements to postpartum metabolic status. Here, a one-off measurement of PRL at 27-30 weeks was positively associated with subsequent maternal beta-cell function at 3 months postpartum, including after adjustment for lactation status.

Discussion
To our knowledge, this is the first systematic review to synthesise the evidence examining PRL in pregnancy and postpartum in relation to maternal metabolic and glycaemic outcomes, including GDM. Systematic reviews addressing mechanistic questions are relatively under-utilised in the endocrine literature, yet are key to assembling disparate data and setting future research agendas. Our results show no clear relationship between PRL and GDM status in the second half of pregnancy (following GDM development), but highlight the lack of evidence regarding the metabolic associations of PRL in early pregnancy. In the postpartum context, particularly with lactation, physiological PRL elevation is associated with low circulating insulin levels, low beta-cell function and insulin sensitivity; although the direction of causality remains unclear.

Prolactin in pregnancy metabolism and GDM aetiology
Pre-clinical evidence prior to this review provides strong theoretical support for the role of PRL in GDM aetiology. Hence, studies examining PRL in relation to GDM or maternal metabolic parameters during pregnancy comprised a large proportion of our review (13 of the 26 included studies). Among these studies, examination of PRL in early pregnancy (prior to GDM development) was a relatively uncommon approach. The largest of these studies prospectively linked higher first-trimester PRL levels to an increased risk of developing GDM (21), but this finding was not corroborated in the remaining studies (15,23,24). In studies focusing on PRL levels in late pregnancy (>24 weeks, at or after GDM development/diagnosis), our pooled meta-analysis suggested no significant differences between late pregnancy PRL in women with GDM compared with controls. Our findings thus suggest that late pregnancy PRL is likely not associated with GDM, but the contribution of earlier PRL concentrations and/or temporal variations in PRL to the development of GDM remains unknown.
These findings are interesting, given the growing body of general evidence in support of metabolic actions for PRL. In non-pregnant populations, observational evidence suggests that the effects of circulating PRL concentrations on metabolism and glucose homeostasis may be concentrationdependent (21). Large, population-based observational studies in non-pregnant, middle aged adults consistently suggest that higher PRL levels within the normal physiological range may be protective against the development of T2DM (40-44), a finding confirmed in a recent systematic review (45) (but not demonstrated here in the context of GDM). However, in states of pathological hyperprolactinaemia (such as in patients with prolactinoma, or treated with anti-psychotic medications), PRL levels well in excess of the normal range have been repeatedly associated with adverse metabolic outcomes including hyperinsulinaemia, insulin resistance, impaired endothelial function, elevated inflammatory markers and increased body weight (46-48). Such outcomes may be ameliorated by treatment of the pathological PRL elevation (with bromocriptine, for example) (47,(49)(50)(51). The contribution of PRL elevation to the hormonally-mediated insulin resistance of pregnancy is debated, but may again be dose-dependent: whilst low levels of PRL have been shown to inhibit lipolysis, the higher concentrations associated with late gestation have traditionally been thought to contribute to progressive stimulation of lipolysis and reduced insulin sensitivity (3,52). Furthermore, other research, largely conducted in animal models or in vitro, also provides strong theoretical support for the parallel role of PRL (alongside that of hPL) in promoting maternal islet cell adaptation and proliferation, a key adaptation to the increased insulin requirements of human pregnancy (6,53). The increase in maternal insulin secretion during pregnancy is paralleled by increasing expression of PRL receptors on maternal pancreatic beta-cells, which bind PRL as well as hPL (52). This concept was established in pre-clinical rodent models, which consistently demonstrated marked increases in beta-cell proliferation and survival, insulin gene expression, and glucoseinduced insulin secretion in response to both hormones in vitro and in vivo (6,53). Indeed, knockout mice specifically lacking PRL receptors on pancreatic beta-cells have normal glucose tolerance outside of pregnancy, but become progressively glucose intolerant with gestation due to corresponding failure of beta-cell proliferationessentially, developing GDM (54, 55). The direct applicability of these animal models to human betacell adaptation in pregnancy remains uncertain, although autopsy studies confirm increased beta-cell mass in pregnant women (56) and in vitro evidence shows that PRL (and hPL) directly enhance insulin secretion from human islets (6).
The lack of a clear relationship between PRL and GDM status in our review (despite the plausible mechanisms, outlined above, that emerge from non-pregnant humans and/or animal models) highlights the likely increased complexity of the interplay between PRL and maternal metabolism in human gestation. The hormonal milieu of pregnancy is multifaceted and synergistic, and several changes independent of PRL may modify insulin secretion and oppose insulin action in peripheral tissues. Rising levels of placental growth hormone, maternal insulin-like growth factor-1 (IGF-1), progesterone, tumor necrosis factor alpha (TNF-a) and cortisol, and a reduction in adiponectin; are also major contributors to progressive insulin resistance in late gestation; such that measurements of a single hormone in isolation are inherently problematic. Moreover, circulating serum levels provide only a partial description of hormone actions: for instance, recent evidence suggests that certain PRL receptor polymorphisms may predict GDM risk, implying differences at a cellular receptor level which may be just as important as absolute circulating hormone concentrations (57).

Prolactin and maternal metabolism in lactation and the postpartum period
In the postpartum period, physiological hyperprolactinaemia is the key endocrine change responsible for the initiation and maintenance of lactation. Lactation is a unique metabolic state associated with an elevation of plasma free fatty acids, and with the mobilisation of lipids from diet and adipose stores to the breast for milk production. Observational evidence suggests that lactation is associated with maternal metabolic benefits, with consistent findings of lower rates of persistent postpartum dysglycaemia and progression to T2DM in women who breastfeed compared with those who do not (both in the general population (58) and following GDM pregnancy (59)). Plausibly, then, PRLas the central 'breastfeeding hormone'may link effective and sustained lactogenesis to improved maternal metabolic status postpartum. Whether this is primarily mediated by improved beta-cell function or reduced insulin resistance remains unclear, as there are putative biological mechanisms for both (3,34,39).
Furthermore, adverse maternal metabolic environments may also have detrimental impacts on lactation success: observational evidence clearly demonstrates that women with obesity and/or diabetes are at significantly increased risk of lactogenesis delay and persistent poor milk supply (60, 61). Some of the studies in our review attempted to link PRL levels postpartum to maternal GDM status, none showing significant relationships: however, maternal lactation status was inconsistent, sample sizes were small and PRL sampling methodology frequently failed to account for the complex and pulsatile PRL dynamics that occur during breastfeeding.
The key findings from our review, however, emerged from those studies that measured PRL as part of a broader examination of the unique endocrine and metabolic environment of lactation. The studies included represent a small subset of a larger body of literature addressing the impact of breastfeeding on maternal metabolism: only studies that measured PRL and then related it directly to a maternal metabolic variable met our criteria for inclusion. Broadly, results support the concept of lactation (under the control of PRL) as a metabolic environment characterised by low circulating insulin levels, increased insulin sensitivity and low beta-cell function (34,36,37). Increased glucose concentrations post glucose load in the presence of low circulating insulin levels in healthy lactating women ensure the availability of glucose for lactose synthesis in the breast, which is not an insulin-dependent process (34). As such, the onset of lactation requires significant changes to glucose metabolism, beginning with beta-cell mass contraction, decreased insulin secretion and reduced peripheral insulin resistance. Lactation is also characterised by increased rates of maternal lipolysis and endogenous glucose production in the fasting state (when dietary substrate is unavailable for milk production) (37). However, the endocrine milieu of lactation is complex; and so the direct hormonal contribution of PRL to these metabolic changes is difficult to ascertain. Furthermore, the directionality of the relationship is unclear: does lactation (under the chief control of PRL) mediate improved postpartum metabolic outcomes, directly contributing to reduced insulin resistance? Or do metabolically-healthy women find it easier to successfully breastfeed, becoming over-represented in the 'lactation' groups in non-randomised observational studies (and having higher PRL simply by virtue of their lactation success)?
Arguments in both directions exist. Clamp data has suggested that (even among lactating women) higher PRL values may enhance insulin-mediated suppression of lipolysis, and lower both intra-hepatic and circulating triglycerides (37). The authors of this work have also previously suggested that lactation may improve postpartum insulin sensitivity by mobilising lipid accumulated in liver and muscle into breastmilk, instead of redirecting lipids into already enlarged adipocytes (3). Conversely, Harreiter et al. (34) have suggested that good "beta-cell plasticity" (i.e. an adequate increase in betacell function during pregnancy and an effective immediate decrease postpartum) is necessary to allow PRL to rise postpartum, exerting a 'permissive' effect on lactation. Indeed, reduced basal PRL levels (31) and reduced PRL responses to infant suckling (38) have been demonstrated in women with overweight/obesity in the studies in our review, and could potentially explain the well-established clinical phenomenon of lactogenesis delay in these individuals. The results of our review suggest that the relationship between PRL secretion and maternal metabolism in the postpartum period is complex, and may be bidirectional, but further data from well-designed and appropriately controlled studies are needed to further clarify these relationships and their implications for maternal and offspring health.

Strengths and limitations
Our review has unique strengths. As mentioned, to our knowledge, it is the first to systematically synthesise the clinical evidence linking PRL to maternal metabolic outcomes in human pregnancy and postpartum. It addresses a unique, mechanistic question linking metabolic and reproductive aspects of women's health; and has allowed us to reach evidence-based conclusions and identify areas for future research.
Limitations of the review process included restricting the search to published work and to the English language.
Limitations of the collated literature included the relatively small number of eligible studies and the marked heterogeneity therein, which precluded meta-analysis for most outcomes. Variable study quality was reflected in the risk of bias assessments (20 of 26 [77%] were deemed to have 'moderate' or 'high' risk of bias). Studies were all observational, with small participant cohorts. In GDM studies specifically, PRL was often only sampled at a single late-pregnancy timepoint (most commonly after the development of GDM) and then compared between GDM and control groups. PRL levels start rising from the beginning of pregnancy and increase rapidly across the latter half of pregnancy, but sampling of PRL within a broad gestational age bracket (eg. 24-28 weeks) was a common approach, often without subsequent adjustment for exact gestational age at the time of collection. Furthermore, PRL exhibits significant diurnal rhythms (both inside and outside of pregnancy) (62), but timing and conditions of collection were unclear and/or unstandardised in many studies. Finally, the diagnostic criteria used to define GDM was inconsistent between studies (see Table 1), reflecting differences in their era of publication and region of origin.
In postpartum studies, methodology failed to acknowledge the complex dynamics of postpartum PRL secretion. Many studies compared one-off measurements of PRL amongst women within broad postpartum timeframes (e.g. 3-5 months postpartum), without detailed description of lactation status and intensity, presence/absence of supplemental feeding, or careful timing of sample collection relative to a feed.
PRL assay methodology also varied according to study age, with older studies using radioimmunoassay techniques and newer studies favouring enzyme-linked immunoassays. Finally, the hormonal environment of pregnancy and postpartum is complex, and studies that focus on absolute serum levels of a single hormone inevitably overlook other factors such as local tissue availability, hormone synergy, and receptor polymorphisms.

Conclusion
In summary, our findings suggest that whilst many human observational studies have attempted to link PRL to GDM pathophysiology, the available evidence is methodologically diverse and conflicting. Overall, there was no clear relationship between maternal PRL levels in late pregnancy and GDM status, while relationships with early pregnancy PRL (preceding GDM development), have not been established and await further study. There were no clear associations between PRL and other maternal glycaemic or weight-related parameters. In the postpartum state, particularly in the context of lactation, a high PRL environment is associated with low circulating insulin levels, low beta-cell function and increased insulin sensitivity. The exact contribution of PRL to these metabolic adaptations remains unclear and warrants further exploration.

Data availability statement
The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding author.