Paradigm Shifts in the Management of Diabetes in Pregnancy: The Importance of Type 2 Diabetes and Early Hyperglycemia in Pregnancy: The 2020 Norbert Freinkel Award Lecture

For over 50 years, the diagnosis of gestational diabetes mellitus (GDM) has been based upon an oral glucose tolerance test (OGTT) at 24–28 weeks’ gestation (1). This time in gestation is when insulin resistance is increasing with the development of hyperglycemia among those with insufficient insulin secretory capacity to maintain euglycemia (2). In Freinkel’s iconic Banting Lecture in 1980, building upon the Pederson hypothesis (3, 4), hyperglycemia was described as causing short- and long-term harm to the growing fetus through “fuel-mediated teratogenesis” (5). Different time-dependent impacts of hyperglycemia on the fetus, shown in Fig. 1, include first trimester effects on organs (leading to congenital malformations), effects on brain development (with behavioral ramifications) across the trimesters, and anthropometric and metabolic effects from the late second trimester onward. At that time, GDM was considered usually to commence at 24–28 weeks, type 2 diabetes in pregnancy was uncommon, and the importance of obesity during pregnancy was largely undescribed. In 1979, 1984, and 1990, three International Workshop Conferences on GDM occurred, “cementing in” approaches to screening (risk factors, then 50-g glucose challenge test [GCT]), diagnosing (100-g, 3-h oral glucose tolerance test [OGTT]), and managing GDM (nutritional counseling; glucose monitoring with targets of <105 mg/dL [5.8 mmol/L] and <120 mg/dL [6.7 mmol/L] pre-/postmeals; insulin therapy if diet “fails”; fetal surveillance and postpartum testing for dysglycemia using a 75-g OGTT) (6).

It is now 40 years since Freinkel’s lecture. This 2020 lecture reflects on the changes and paradigm shifts that have occurred since that time and proposes, in particular, the recognition that hyperglycemia is often present early in pregnancy and that the effects of maternal hyperglycemia at conception, versus those at 24–28 weeks in pregnancy, on fetal development need to be assessed separately. GDM may be divided into mild hyperglycemia already present at the beginning of pregnancy (“prevalent GDM”) and that arising de novo during pregnancy (“incident GDM”). Such recognition should lead to further major changes in our approaches to treating and managing GDM.

In the 1980s, most women with pregestational diabetes had type 1 diabetes. Evidence for a high prevalence of type 2 diabetes among women of reproductive age of non-European descent emerged from studies among indigenous and Pacific populations from the 1960s (7) to other non-European populations, including South Asians (8). Inevitably, growing reports emerged of significant numbers of women with type 2 diabetes in pregnancy (9). For example, between 1994 and 2004 across the U.S., the nationwide prevalence of type 1 diabetes in pregnancy increased from 0.24% to 0.33%, an increase of ∼33% (10). At the same time, the prevalence of type 2 diabetes in pregnancy increased from 0.09% to 0.42%, an increase of ∼367% (10). This led to a shift in pregestational diabetes from a preponderance (73%) of women with type 1 diabetes to a majority (56%) with type 2 diabetes. The growing presence of type 2 diabetes in pregnancy is not simply a case of “mild diabetes.” Pregnancies complicated by type 2 diabetes often have comparable rates of congenital malformations, stillbirths, and other adverse pregnancy outcomes (9), requiring the same degree of obstetric monitoring (and intervention) as pregnancies complicated by type 1 diabetes. There are also additional risks from fetal exposure to potentially harmful antihyperglycemic agents, along with a greater prevalence of actually or potentially teratogenic pharmacological agents to manage components of the metabolic syndrome (e.g., antihypertensives and antilipid agents) (11). There are maternal threats from the progression of retinopathy, nephropathy, and cardiovascular disease.

The management of women with type 2 diabetes in pregnancy is often complicated by severe insulin resistance. While this can reflect coexisting obesity, there is often a need for substantial quantities of insulin. Metformin can help reduce insulin requirements in type 2 diabetes (12) (e.g., from 155.29 ± 134.01 units/day to 109.76 ± 105.1 units/day), but the insulin dose remains sizeable and the broad variance reflects some women needing a substantially higher dose. A small 1:2 case-control study (13) in South Auckland, New Zealand (n = 90, 93% Polynesian) was the first to use continuous subcutaneous insulin infusion (CSII) to overcome the insulin resistance–associated hyperglycemia for women with either preexisting type 2 diabetes (33%) or GDM requiring >100 units/day of insulin (69% >200 units/day). The women on CSII required a median (range) of 246 (116–501) units/day or 2.68 units/kg booking (first antenatal) weight. Glycemia had improved among 79% of the women within 1–2 weeks. Outcomes were similar in terms of birth weight (3,790 vs. 3,720 g), proportion of babies <3,000 g (10.2% vs. 11.8%), and babies ≥4,000 g (34.4% vs. 40.0%). While other obstetric outcomes were comparable, the proportion admitted to the special care baby unit (56.3% vs. 25.0%) was higher in those treated with CSII. No reason is given for these excess admissions, and this finding may be a flag that we need to be “balanced” in our efforts to reduce fetal exposure to hyperglycemia.

In the Metformin in Women With Type 2 Diabetes in Pregnancy Trial (MiTy) (12), metformin therapy was associated with a much lower insulin requirement, alongside a 0.18% lower “last HbA_1c in pregnancy,” 0.1 kg lower weekly weight gain, and 1.77 kg less overall weight gain. This mix of metformin therapy, less weight gain, and less hyperglycemia (the mean last HbA_1c [5.9%] was still higher than the pregnancy HbA_1c target of 5.6% [14]) was associated with 0.59-fold (95% CI 0.36–0.98) less extreme large-for-gestational-age (LGA) babies born but a 2.07-fold (1.16–3.71) greater proportion of small-for-gestational-age (SGA) babies. While the need for high-level neonatal care >24 h was comparable, the need to “balance” diabetes management with avoiding fetal undernutrition (and the potential for long-term metabolic risks [15]) clearly requires further research. The mechanism behind the SGA remains unclear, with potential contributions from metformin exposure, less gestational weight gain, more mild hypoglycemia episodes/week, and current early pregnancy glycemic targets.

Of course, type 2 diabetes often remains undiagnosed. This growing problem was fully acknowledged by the International Association of the Diabetes and Pregnancy Study Groups (IADPSG) in their recommendations for the new diagnostic classification for hyperglycemia in pregnancy (16). The IADPSG divided GDM into two on clinical grounds, with likely undiagnosed type 2 diabetes as “overt diabetes in pregnancy” (ODIP) and GDM as “hyperglycemia first detected in pregnancy less than overt diabetes.” This classification was subsequently adopted by the World Health Organization, changing its name to “diabetes in pregnancy” (17). This clinical classification thereby allows triaging into those who require, e.g., retinal and renal screening and more intensive management (women with ODIP) and those requiring more limited management (GDM). Although unrelated to the process to define criteria for GDM from the Hyperglycemia and Adverse Pregnancy Outcomes (HAPO) OGTT results at 24–28 weeks (18), this new clinical classification addressed the important paradigm shift in clinical care that had occurred with the growing numbers of women of reproductive age with diagnosed and undiagnosed type 2 diabetes at conception.

The IADPSG recommended that ODIP be diagnosed using the criteria for diabetes in nonpregnant adults, with either a fasting glucose (FBG) ≥7.0 mmol/L (126 mg/dL), HbA_1c ≥6.5% (48 mmol/mol), and/or a random glucose (RBG) (with confirmation) ≥11.1 mmol/L (200 mg/dL) (16). Pregnancy complications of ODIP are indeed increased (19). One key finding is that not all women with ODIP are found to have type 2 diabetes postpartum (20). Whether this is due to their antenatal lifestyle changes and other aspects of diabetes in pregnancy management or simply an issue with glycemic variance is unclear.

Screening for ODIP can have its own challenges, and this was exposed by the social distancing required during the coronavirus disease 2019 (COVID-19) pandemic (21, 22). Attending for an FBG can require many women to visit a pathology collection center concurrently. To space out attendance through the day, HbA_1c and/or RBG have been recommended, but their test characteristics demonstrate their limitations. In a retrospective analysis of 17,852 pregnancies (19), with a universal RBG around the initial antenatal visit, the area under the receiver operator curve for RBG to detect ODIP was 0.86 (0.80–0.92), validating its use as a screening test in that population. Testing for ODIP using HbA_1c in early pregnancy has good reproducibility, but its utility is influenced by the variability in half-life between individuals and other nonglycemic factors (23). Even the 6.5% (48 mmol/mol) threshold is only 50% sensitive for ODIP by OGTT (24). For screening for ODIP, a threshold of 5.9% (41 mmol/mol) (24) is associated with two- to fourfold increased risk of adverse pregnancy outcomes, but test characteristics have not been shown. A slightly lower threshold of 5.7% (39 mmol/mol), as used by the American Diabetes Association for prediabetes, is a poor predictor of adverse pregnancy outcomes among obese European women when there is little ODIP (25) (e.g., adjusted odds ratios for birth weight >4 kg, preeclampsia, and cesarean section were 0.94, 0.77, and 0.55, respectively). The OGTT therefore remains the best test for hyperglycemia in pregnancy as it detects both fasting and postprandial hyperglycemia, although reproducibility, convenience, and potentially social distancing issues remain. Further studies are needed to identify the best screening regimen early in pregnancy if the OGTT is not to be used.

Besides the question of how best to test for ODIP, the shift to screening for undiagnosed diabetes has revealed another issue that has been simmering away since Norbert Freinkel’s time. In 1988, Maureen Harris asked whether GDM was simply preexisting impaired glucose tolerance rather than newly developed hyperglycemia (26). The potential for GDM (that is not ODIP) to be prevalent in the first half of pregnancy, and to be of importance, was not reflected in Freinkel’s iconic 1980 figure.

The existence of hyperglycemia less than ODIP (GDM) early in pregnancy raises a multitude of questions:

• Is the etiological mechanism for GDM from early pregnancy the same as GDM developing later on?

• Is there a difference in the impact of GDM developing earlier and later on pregnancy outcomes and fetal programming?

• What tests and diagnostic thresholds should be used to “diagnose” GDM in early pregnancy? Should different criteria be used early and later in pregnancy?

• Is the treatment of GDM detected in early pregnancy of benefit? Could associated early medication, lower glucose levels including hypoglycemia, and/or insufficient weight gain cause harm?

Depending on the setting, screening approach, and diagnostic criteria used, a high proportion (15–70%) of women with GDM have hyperglycemia detectable in early pregnancy (27). Women who develop GDM in early pregnancy appear to be more insulin resistant, with greater waist circumference, higher blood pressure, and higher triglycerides (28), than women who develop GDM later in pregnancy (29). Most importantly, a meta-analysis of 13 cohort studies showed that perinatal mortality (relative risk [RR] 3.58 [1.91–6.71]), neonatal hypoglycemia (RR 1.61 [1.02–2.55]), and insulin use (RR 1.71 [1.45–2.03]) were greater among women with early GDM compared with those with GDM diagnosed later in the pregnancy (27). This was despite their GDM receiving standard treatment. The question arises why outcomes were worse in spite of treatment: longer/greater exposure to hyperglycemia and/or similar pathophysiology but with greater severity, and/or a different etiological mechanism?

The proposition that GDM is heterogenous (including early vs. late) is important, as it may suggest different GDM preventive and treatment strategies for different types of GDM. Besides the small numbers of those progressing to type 1 diabetes, monogenic “diabetes” (including maturity-onset diabetes of the young [MODY] 2), and forms associated with other genetic conditions, GDM is proposed to be associated with three different phenotypes (by definition, all have insufficient insulin secretion for their needs) (30). These include reduced insulin sensitivity with normal insulin secretion, normal insulin sensitivity with reduced insulin secretion alone, and reduced insulin sensitivity with reduced insulin secretion (30). The initial studies were in late pregnancy, but the same heterogeneity is shown in early pregnancy, with the reduced insulin sensitivity forms predominant (31). There may be differences in pregnancy outcomes with the less-insulin-sensitive being at greater risk of LGA and cesarean section (30, 31). Perhaps management should differ between these phenotypes as well as between GDM diagnosed earlier and later in pregnancy.

Prior to HAPO and the development of the IADPSG GDM diagnostic criteria, criteria were based upon prediction of future maternal type 2 diabetes (1), complications among nonpregnant adults (World Health Organization), or “glycemic distribution” (32). The IADPSG process took the three-point, 2-h, 75-g OGTT data from the HAPO study and gained broad (but, alas, incomplete) agreement over GDM diagnostic thresholds based upon the linear relationship between glucose and adverse pregnancy outcomes (16). An odds ratio of 1.75-fold risk of adverse outcomes at each of the three time points using the HAPO data was chosen. This approach to equilibrate adverse pregnancy outcome risk at different OGTT time points revealed discordance between existing and proposed fasting (needed to be lower) and 2-h (needed to be higher) glucose thresholds on the 75-g OGTT (16). As HAPO did not use a screening test, the 50-g GCT itself (having little empirical evidence [33]) was dropped. As a result, many of those with an isolated high fasting glucose that had previously been excluded from the OGTT were suddenly being diagnosed with GDM. The IADPSG criteria also require only one, rather than two, time points to be above the threshold, further increasing the importance of the fasting glucose threshold. Overall, 50% of HAPO women were diagnosed on the fasting result: a shift in the GDM glycemic profile. However, applying the IADPSG criteria to the HAPO results also revealed significant heterogeneity in the OGTT pattern, particularly between ethnic groups, some of whom were considered hyperglycemic predominantly based upon the 1-h and/or 2-h post–glucose load test results (34).

However, HAPO tested women at 24–28 weeks, not in early pregnancy. A similar study involving 6,129 women but using only a fasting glucose test at a median of 9.5 weeks showed a similarly linear increase in pregnancy complications as glucose increased from <75 mg/dL (4.2 mmol/L) to 100–105 mg/dL (5.6–5.8 mmol/L) (35). The adverse pregnancy outcomes LGA/macrosomia and primary cesarean section increased from 7.9% to 19.4% and from 12.7% to 20.0%, respectively, across the glucose categories. GDM at 24–28 weeks using the 50-g GCT and 100-g OGTT increased from the lowest to the highest glucose category from 1% to 11.7%, respectively. Unlike HAPO, this was not a blinded study, and hence women were presumably treated. These data contributed to a growing shift in paradigm that women with higher glucose levels early in pregnancy, below overt diabetes, should be treated for their GDM.

Before the IADPSG recommendations for managing hyperglycemia in pregnancy, GDM was generally diagnosed early in pregnancy by applying existing GDM criteria used at 24–28 weeks. Overdiagnosis was avoided by limiting early OGTTs to those women with a prior history of GDM or, e.g., glycosuria. However, with the introduction of screening for ODIP among women with other type 2 diabetes risk factors, more women were found to have GDM using these criteria. The IADPSG acknowledged that there were limited trial data and initially recommended that an FPG ≥5.1 mmol/L (92 mg/dL) in early pregnancy also be used to diagnose GDM (16). However, within 5 years, data emerging from Italy and China (36) suggested many such women no longer fulfilled GDM when later retested. The Chinese study (37) found that at 24–28 weeks’ gestation, GDM was only present in 37.0%, 52.7%, and 66.2%, respectively, of women with an FPG at the first antenatal visit of 5.10–5.59, 5.60–6.09, and 6.10–6.99 mmol/L. In response, the IADPSG recommended using an HbA_1c, and not a fasting glucose, to test for hyperglycemia in pregnancy (37). The situation is compounded by the potential for the “fetal steal phenomenon” (38) to have lowered maternal glycemia, with the offspring already hyperinsulinemic and at risk for macrosomia. Furthermore, the HbA_1c has now been shown to be an insensitive test for both hyperglycemia and adverse pregnancy outcomes (25), and two randomized controlled trials (RCTs) of treating women with HbA_1c 5.7–6.4% (39, 40) showed no treatment benefit.

One major RCT has now been completed of early testing for GDM among 962 obese women with a subgroup analysis among those who had GDM diagnosed (early screening n = 69 [15.0%] vs. routine screening n = 56 [12.1%]) (41). Screening involved the two-step approach (50-g GCT followed by 100-g 3-h OGTT). Randomization was at mean 13.6–13.8 weeks’ gestation. There was no significant difference in any pregnancy outcome, although insulin treatment was 3.70-fold (1.04–13.17) more likely in the early-screen group. In fact, the primary composite outcome (macrosomia, primary cesarean delivery, gestational hypertension, preeclampsia, hyperbilirubinemia, shoulder dystocia, and neonatal hypoglycemia) was nonsignificantly higher in the early-screen group (56.9% vs. 50.8%, P = 0.06) with preeclampsia (13.6% vs. 9.5%, P = 0.06), in particular, nonsignificantly higher. Gestational age at delivery was lower (36.7 vs. 38.7 weeks, P = 0.001) among those with GDM in the early-treatment group, but the groups were too small to show a difference in less common, but more severe, outcomes. The early-screen group of course included treating the ∼2.9% of women who would not have had GDM diagnosed at 24–28 weeks’ gestation. Similarly, no significant difference in adverse pregnancy outcomes was found between women randomized to either lifestyle intervention (n = 36) or control (n = 54) in a Danish GDM prevention study post hoc analysis of participants with a fasting glucose 5.1–6.9 mmol/L and a 2-h capillary blood glucose <9.0 mmol/L (42).

Managing GDM is not without risk. While the iconic work of Freinkel and Pederson emphasized short- and long-term risks from fetal fuel oversupply (5), the “thrifty phenotype hypothesis” postulates that fetal undernutrition leads to future metabolic risks (15). Antenatal metformin therapy, insufficient gestational weight gain, and overmedication are all potentially associated with disturbances in fetal metabolism that might therefore lead to future obesity and dysglycemia (15, 43, 44). The TOBOGM RCT was commenced to address the equipoise between the potential for harm and there being no existing major RCTs showing benefits from treating GDM at the time of “booking” (i.e., early in pregnancy) (45, 46). There is good RCT evidence to show that GDM treatment at 24–28 weeks is of benefit (47, 48), although neither major trial used the IADPSG criteria.

The TOBOGM study is a multicenter RCT testing whether diagnosing and treating GDM from booking, rather than waiting for the OGTT at 24–28 weeks, reduces adverse pregnancy outcomes. Women at <20 weeks’ gestation with GDM/diabetes risk factors are invited for an early OGTT. With this early OGTT, women are identified with “booking GDM” using the IADPSG 24–28 weeks criteria and randomized either to immediate referral for GDM management (and no further OGTT) or to a repeat OGTT at 24–28 weeks’ gestation. Clinicians are blinded to OGTT results. Antenatal and GDM care otherwise follow local/agreed guidelines. A pilot study (46) has been completed and the main TOBOGM RCT is currently underway (47). The pilot RCT randomized 11 women with “booking GDM” to early treatment and 10 women to a repeat OGTT at 24–28 weeks’ gestation. A further 58 women served as decoys (i.e., women with normal glucose tolerance mingling in clinic with and labeled in the same way as the women randomized to untreated “booking GDM”). Important findings from the TOBOGM pilot study were, first, that GDM was still present in 89% of the women randomized to no early treatment and had developed in 20% of decoys. However, most important was that neonatal intensive care unit admission was highest in the treated group (36% vs. 0%, P = 0.043), largely due to SGA babies, while LGA babies were more common in the no-early-treatment group (0% vs. 33%, P = 0.030). This may indicate that early treatment (including gestational weight gain limitation) may have both benefits and harms, making completion of the main TOBOGM RCT even more important. It certainly appears that lifestyle intervention (49) and weight management (50) have different impacts on GDM risk earlier and later in pregnancy and may require different management approaches early and later in pregnancy.

The main multicenter RCT is currently underway, requiring approximately 4,000 women to be recruited to identify 800 women with “booking GDM” with 400 randomized to early GDM treatment and 400 to the second OGTT at 24–28 weeks’ gestation. A further 800 randomly selected women serve as decoys, with the remaining 2,400 for chart review only. Randomization is stratified by site. Women with ODIP and those with a fasting glucose 6.1–6.9 mmol/L are excluded from the RCT. The primary pregnancy outcome to detect the effects of maternal hyperglycemia is a composite of respiratory distress, phototherapy, birth trauma, birth <37 weeks, stillbirth/death, shoulder dystocia, and birth weight ≥4.5 kg. The primary neonatal outcome to detect any fetal undernutrition is neonatal lean body mass. The primary maternal outcome is preeclampsia. TOBOGM also stratifies its randomization within two glycemic bands: 1) the HAPO 1.75 odds ratio band (FBG 5.1–5.2 mmol/L and/or 1-h blood glucose 10.0–10.5 mmol/L and/or 2-h blood glucose 8.5–8.9 mmol/L), and 2) the HAPO 2.0 odds ratio band (FBG 5.3–6.9 mmol/L and/or 1-h blood glucose ≥10.6 mmol/L and/or 2-h blood glucose 9.0–11.0 mmol/L). The analyses will not only compare pregnancy outcomes between early-treated women and those who had the deferred OGTT but will repeat the comparisons within the two glycemic bands and between these bands and the decoys. Further analyses will be undertaken to identify the optimal OGTT time point thresholds.

TOBOGM is the largest and most comprehensive RCT comparing treatment for women with early GDM and those with treatment deferred to the outcome of the OGTT at 24–28 weeks, and results will likely not be available for 2–3 years. In the meantime, the IADPSG recommendations provide no guidance for where to diagnose GDM <7.0 mmol/L and/or 2-h glucose <11.1 mmol/L (with confirmation). The American Diabetes Association recommends those with “prediabetes” be treated as if they have hyperglycemia (fasting ≥5.6 mmol/L and/or 2-h glucose ≥7.8 mmol/L), but there may be a risk of overtreatment early in pregnancy. HbA_1c testing has been advised as an alternative, but there is no evidence of benefit. In terms of diagnosis, a 75-g OGTT is probably still the best test (with all its challenges). Previously, we have recommended using a threshold of 6.1–6.9 mmol/L to diagnose GDM early in pregnancy and this agrees with the recommendations by Zhu et al. (27, 37). A 2-h threshold of ≥8.5 mmol/L would identify those with significant postprandial hyperglycemia. All women without “early hyperglycemia in pregnancy” would then undertake the OGTT at 24–28 weeks.

There have been several paradigm shifts since Freinkel’s iconic lecture of 1980. These include the growth in the importance and our understanding of type 2 diabetes in pregnancy, insights into the role of maternal obesity on fetal growth and metabolism, the realization that GDM is heterogenous with hyperglycemia often already present at conception, the introduction of oral antihyperglycemic agents in therapy, and the importance of the first trimester in the growth trajectory and metabolism of both the mother and offspring. Figure 2 brings each of these concepts into an adaptation of Fig. 1, also acknowledging the importance of the setting in which the pregnancy occurs.

Future work needs to consider the consequences of this new-found variation. This includes the potential for different screening strategies and diagnostic criteria early and later in pregnancy. Similarly, the use of metformin, weight management, glucose treatment thresholds, and lifestyle strategies may need to differ between women with GDM diagnosed early in pregnancy (who appear to have a more insulin-resistant pattern) and those diagnosed later. In view of these observed differences, it may be necessary to divide GDM into mild hyperglycemia already present at the beginning of pregnancy (“prevalent GDM”) and that which arises de novo during pregnancy (“incident GDM”). Perhaps the move to subclassify GDM in this way should start now.

Acknowledgments. I am especially grateful to my mentors in epidemiology, Rhys Wlliams from Cardiff, Wales, and Paul Zimmett, Melbourne, Australia; my mentor in diabetes in pregnancy, the late David J. Scott; and my mentor in academic medicine, the late P. John Scott, who have between them guided me to broader and deeper insights into research. I am also grateful to my cultural mentors, all now passed, who gave me insights into indigenous communities: Betty Hunapo, Ngati Hine, Aotearoa/New Zealand; Buddy Te Whare, Ngati Maniapoto, Aotearoa/New Zealand; Rick Henderson, Goulburn Valley, Australia. I am grateful to all of the teams I work/have worked with in South Auckland and the Waikato in New Zealand, in Victoria and New South Wales in Australia, and the Addenbrookes/Rosie teams in Cambridge. Many thanks to the Vitamin D and Lifestyle Intervention for GDM Prevention (DALI) team for all of their insights—what a great group! Also thanks to my Swedish team members in Örebro, from whom I have learned a lot about obstetrics! I have to thank my colleagues in the Diabetic Pregnancy Study Group (DPSG) and Australasian Diabetes in Pregnancy Society (ADIPS) who have given such terrific insights into hyperglycemia in pregnancy over the years. I also wish to thank my good colleague Gernot Desoye, University of Graz, Graz, Austria, for critically reading the manuscript and important input into Fig. 2. Many thanks to Anand Hardiker for the artwork and assistance with Fig. 2.

Funding. This work has been supported by multiple funders, particularly the Health Research Council in New Zealand, the European Commission, and the National Health and Medical Research Council of Australia.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.