Falling Insulin Requirements Are Associated With Adverse Obstetric Outcomes in Women With Preexisting Diabetes

Falling insulin requirements in late pregnancy are thought to signify feto-placental compromise, often prompting admission to hospital for maternal and fetal monitoring as well as emergency delivery in some cases (1). Most women with preexisting diabetes experience an increase in insulin requirements during the second half of pregnancy (2) to maintain euglycemia, in light of a 50–70% rise in insulin resistance (3). Because insulin resistance is driven by placental-mediated hormone production, including human placental lactogen (hPL), cortisol, and progesterone (4,5), a fall in insulin requirements has been interpreted as evidence of placental dysfunction, and often precipitates clinical concern and possible management intervention. However, there are very few published data that ascertain the significance of a large fall in insulin requirement and its significance for maternal and fetal well-being.

The limited cohort studies examining this group of patients have not shown any difference in outcomes (6–8). Furthermore, no previous studies have directly associated falling insulin requirements with evidence of placental insufficiency, and it remains unclear why some women have large falls in insulin requirements and what the underlying pathophysiology is. Importantly, previous studies have not included women with type 2 diabetes, which is now more prevalent than type 1 diabetes in women entering pregnancy (9). Knowledge of the clinical significance of falling insulin requirements is needed to help guide clinical management and avoid unnecessary intervention. In view of this paucity of data, we conducted a retrospective review to investigate the clinical significance of falling insulin requirements in late pregnancy in women with preexisting type 1, type 2, or overt diabetes in pregnancy.

We conducted a chart review of women delivering between 1 January 2010 and 31 January 2013 who were managed in the Diabetes in Pregnancy antenatal clinic of two hospitals in Sydney, Australia. Between them, these hospitals manage 7,500 deliveries annually. The study was approved by the Western Sydney Local Health District ethics committee.

All women with a diagnosis of preexisting type 1 or type 2 diabetes were included. In addition, women diagnosed with “overt diabetes” on a 75-g oral glucose tolerance test during pregnancy were included. Overt diabetes was defined as a fasting blood glucose level of ≥7.0 mmol/L (126 mg/dL), a 2-h level of ≥11.1 mmol/L (200 mg/dL), or an HbA_1c of ≥6.5% (48 mmol/mol), consistent with the International Association of Diabetes and Pregnancy Study Groups’ definition (IADPSG) (10). This definition of overt diabetes was applied retrospectively at the time of chart review. During pregnancy, they were managed under a glucose control protocol, which was common both to women with preexisting diabetes and to women with gestational diabetes.

Information on patient demographics, history of diabetes, and pregnancy outcomes was obtained from the patient notes and a statewide electronic obstetric database called “Obstetrix.” Information for this database was entered prospectively by clinic staff and included data on patient demographics, obstetric and medical history, current pregnancy management, details of the delivery, and maternal and neonatal outcomes. Insulin dosage, weight in kilograms, and blood pressure were recorded at each review. Both institutions follow similar clinical practice with women reviewed at least every 2–4 weeks until 28 weeks’ gestation and every 1–2 weeks thereafter until delivery. Insulin dose was titrated to the same blood glucose target of 5.5 mmol (99 mg/dL) fasting and 7 mmol (126 mg/dL) 2 h postprandial, at both hospitals. Daily insulin requirement at each review was recorded as total insulin dose in units (TID, including basal and prandial components) and was corrected for body weight (units/kg). The percent fall in insulin dosage (PFID) was calculated from the peak total insulin dose (PTID), defined as the highest TID in pregnancy, and the trough total insulin dose (TTID), defined as the lowest TID following the peak, using the following formula: PFID = (PTID − TTID)/PTID × 100. Care was taken not to include the period within 5 days of antenatal steroid administration when calculating the PTID.

As there is currently no gold standard definition of placental insufficiency, the primary outcome comprised of a composite of clinical outcomes consistently associated with placental dysfunction in the literature (11–15). These were defined prior to analysis of the data as preeclampsia (systolic blood pressure ≥140 mmHg, diastolic blood pressure ≥90 mmHg, identified on repeated readings several hours apart, and proteinuria ≥300 mg/24 h) (16), small for gestational age (SGA, birth weight ≤5th percentile for gestational age) (17), preterm delivery (≤30 weeks) (18), and placental abruption and stillbirth (>20 weeks). Presence of one or more of these outcomes in the pregnancy was the threshold for attribution of an abnormal primary outcome. Components of the primary outcomes were also examined individually as secondary outcomes.

The secondary outcomes were mode of delivery; maternal “hypertensive disorders of pregnancy,” including gestational hypertension (systolic blood pressure ≥140 mmHg and diastolic blood pressure ≥90 mmHg first identified after 20 weeks of pregnancy, in the absence of proteinuria, which resolved by 12 weeks postpartum) and preeclampsia or eclampsia; and a number of neonatal outcomes (gestation at delivery, preterm delivery ≤30 weeks, admission to the neonatal intensive care unit [NICU], hypoglycemia defined as a blood glucose reading of <2.5 mmol/L [45 mg/dL], jaundice, congenital malformation, birth trauma, 5-min Apgar score <9, documentation of fetal distress, and mortality). Customized neonatal birth weight centiles were calculated, corrected for maternal height, weight, ethnicity, parity, gestational age, and sex of the baby using the Gestation Network Australian centile calculator (19). Birth weight was further classified as large for gestational age (≥95th customized percentile) and SGA (≤5th customized percentile). As the absolute numbers for each neonatal outcome were small, a composite of any of the above neonatal outcomes was used. The umbilical artery systolic-to-diastolic velocity ratio (S/D ratio) on Doppler flow velocimetry was also recorded. A high S/D ratio (defined as ≥95th percentile) was considered to signify abnormal placental function (20).

Analysis was conducted using a case-control model by dichotomizing the cohort into case subjects (defined as women with ≥15% PFID) and control subjects (defined as women with <15% PFID) to determine the predictors and consequences of falling insulin requirements. The cut point of 15% was chosen as this was considered clinically significant in previous studies (6,7). Statistical analysis was performed using SPSS version 20 and S-PLUS version 8. Many of the continuous variables were not normally distributed; therefore the median and lower to upper quartile (Q1–Q3) were used to summarize these distributions, whereas categorical variables were summarized using frequencies and percentages. The Mann-Whitney U and Kruskal-Wallis tests were used to determine differences in the distribution of continuous variables, and Pearson χ² or Fisher exact tests were used to test for association between categorical variables. Odds ratios (ORs) and 95% CIs were calculated using logistic regression analysis to quantify the extent of association between falling insulin requirements and obstetric outcomes. Two-tailed tests with a significance level of 5% were used throughout.

One hundred and sixty-four pregnancies were reviewed from a 3-year period. After exclusion of women with 1) multiple gestation pregnancies (n = 6), 2) pregnancy loss or termination prior to 20 weeks (n = 5), 3) incomplete data on insulin dosage or pregnancy outcome (n = 10), and 4) no insulin treatment required throughout the pregnancy (n = 4), a total of 139 pregnancies in 129 women were included. This consisted of 43 with type 1 diabetes, 84 with type 2 diabetes, and 12 with overt diabetes first diagnosed in pregnancy. Thirty-five pregnancies (25.2%) demonstrated a ≥15% fall in insulin requirements and were defined as case subjects. A comparison of maternal characteristics and insulin requirements during pregnancy between case and control subjects (women without a ≥15% fall in insulin requirements) is summarized in Table 1. Nulliparity was the only baseline factor that was different between case and control subjects. There was no difference in the incidence of falling insulin requirements between women with type 1 and type 2 diabetes (27.9 vs. 20.2%, P = 0.33). However a higher proportion of women with overt diabetes developed falling insulin requirements compared with women with type 2 (50.0 vs. 20.2%, P = 0.034) but not type 1 diabetes (50.0 vs. 27.9%, P = 0.177). A summary of the pattern of insulin requirements in the 10 weeks prior to delivery is shown in Fig. 1. The median fall in insulin requirements in the case group was 27.8% (20.8–42.9), close to double the chosen cutoff of 15%.

Women with falling insulin requirements had an increased risk of the composite of clinical markers of placental dysfunction (34.3 vs. 10.6%, OR 4.4 [95% CI 1.73–11.26], P = 0.002) (Fig. 2A). As the presence of preeclampsia can significantly alter clinical management and outcomes, we conducted a post hoc analysis with exclusion of women who had preeclampsia to ensure that the association of the primary outcome was not attributable solely to this complication. Women with falling insulin requirements continued to have a significantly increased risk of the composite of the remaining clinical markers SGA, preterm delivery (≤30 weeks), placental abruption, and stillbirth (>20 weeks) (17.9 vs 5.2%, OR 4.0 [95% CI 1.1–15.0], P < 0.05) (Fig. 2B).

The maternal, neonatal, and delivery outcomes of women with and without a ≥15% fall in insulin requirements are outlined in Table 2. Babies of women with falling insulin requirements were much more likely to be admitted to NICU, were delivered slightly earlier, and were more likely to be SGA. However, there was no other difference in neonatal outcomes between the two groups, including the composite variable of “any neonatal outcome.” Women with falling insulin requirements had an increased risk of preeclampsia. This association remained significant after adjustment for the potential confounding variables of nulliparity and the presence of baseline proteinuria, by logistic regression (OR 3.47 [95% CI 1.1–11.8], P = 0.045) (Supplementary Table 1). The incidence of other maternal outcomes, mode of delivery, induction of labor, and S/D ratio on ultrasound were similar between the two groups. In addition, weight gain or HbA_1c level during pregnancy was not associated with falling insulin requirements.

As 10 women had repeated pregnancies during the study period, a subanalysis was conducted, including only the index pregnancy in these women (n = 129). Results are available in Supplementary Tables 2–4. Exclusion of the repeated pregnancy did not alter the outcomes. In addition, the percent fall in insulin requirements was highly variable between the first and second pregnancy and could not be predicted by the pattern in the index pregnancy (Supplementary Table 4). Falling insulin requirements remained an independent predictor of preeclampsia after adjustment for nulliparity and baseline proteinuria in this cohort.

The majority of women with preexisting diabetes need increasing doses of insulin with advancing gestation; however, a proportion of women have falling insulin requirements in the latter stages of pregnancy. In our study, one in four women had a fall in insulin requirement of >15% from their peak insulin dose to delivery, with the median reduction in dose close to 30% in this group. We found that women with falling insulin requirements had a significantly greater risk of adverse clinical outcomes associated with placental dysfunction, including preeclampsia. Furthermore, babies of women with falling insulin requirements had significantly higher incidence of NICU admission, were delivered slightly earlier, and were more likely to be SGA.

It is postulated that a reduction in the placentally derived hormones that mediate insulin resistance is responsible for this decrease in insulin requirement in late pregnancy. These hormones include hPL, progesterone, cortisol, leptin, and cytokines such as tumor necrosis factor-α (TNF-α) (4,5,21,22). Although no studies have directly examined these biomarkers with respect to falling insulin requirements, studies have found reduced hPL levels correlate with placental dysfunction (23). As a result, falling insulin requirements in late pregnancy are thought to signify feto-placental compromise and increased obstetric risk. Our study is the first to provide data in support of this theory. Alternative mechanisms for the reduction in insulin requirements in late pregnancy include reduced clearance of exogenous insulin (24), increased endogenous production of insulin and insulin-like hormones (e.g., IGF1) (25,26), and changes in glucose consumption and endogenous production in late pregnancy. However, studies in women with diabetes have been inconsistent and these mechanisms are unlikely to explain the large and sudden fall in insulin requirements seen in some patients. Metformin treatment could also potentially influence insulin sensitivity and the magnitude of change in insulin requirements. In addition, the dose and timing of metformin treatment could precipitate a fall in insulin requirements, independent of placental function. Although 29 (20.8%) women in our cohort were treated with metformin during the pregnancy, we did not find a significant difference between the risk of falling insulin requirements and metformin treatment.

Although there is no gold standard investigation to diagnose placental insufficiency, several previous studies have established the clinical consequences of chronic placental insult (11–15). These include impaired fetal growth, placental separation leading to abruption, pregnancy-induced hypertension, preterm delivery, and stillbirth. Focusing our primary outcome on a composite of these important clinical effects of placental dysfunction is a strength of our study, providing a novel and clinically relevant approach to investigating the association with falling insulin requirements. Our results indicate that one in three women with a ≥15% fall in insulin requirements in late pregnancy had a complication related to placental dysfunction. This translated to a fourfold increased risk of the primary outcome compared with the control group. Furthermore, this association was maintained even after excluding all women with preeclampsia, which in itself influences obstetric management and outcomes (27). The persistence of the association suggests that the adverse outcomes seen with falling insulin requirements are not due to preeclampsia alone.

We are the first to report an association between falling insulin requirements and preeclampsia, with 20% of women in the case group being affected. This was independent of other risk factors for preeclampsia, including nulliparity and presence of proteinuria at baseline. The increased risk of preeclampsia in women with diabetes (28) and the role of the placenta in its pathogenesis are well established (27). Preeclampsia therefore provides insight into the possible mechanisms implicated in the pathogenesis of falling insulin requirements. Animal and human studies suggest that an exaggerated inflammatory response and other factors causing suboptimal placentation in early pregnancy can lead to release of anti-angiogenic factors, causing the preeclampsia syndrome later in pregnancy (29–31). Women with diabetes are more likely to have a hypoxic and proinflammatory intrauterine environment due to underlying vascular disease and particularly in the setting of poor glycemic control (32). The histopathology of preeclamptic placentae reveals infarcts, atherosis, thrombosis, and chronic inflammation leading to a state of placental insufficiency (33). A similar underlying pathology in women with falling insulin requirements, impairing the placenta’s ability to produce hormones mediating insulin resistance, could explain our clinical findings. Thus, it is plausible that falling insulin requirements act as a clinical marker of underlying placental pathology, which lies on the same spectrum of disease as preeclampsia and intrauterine growth restriction (IUGR).

Fetal growth restriction is a well-established consequence of placental insufficiency (11,15). In accordance with this, we found that babies of women with falling insulin requirements were over three times more likely to be SGA (≤5th percentile). Although we acknowledge that this definition of SGA could include some babies that are constitutionally small, it is in keeping with our other clinical findings providing further evidence that the underlying pathology is related to placental dysfunction. As such there is currently no consensus over the definition of IUGR; however, commonly used definitions include estimated fetal weight below the 10th, 5th, or 3rd percentile for gestational age, consistent with our methods (17). The utility of umbilical artery Doppler has been demonstrated in pregnancy complicated by IUGR and preeclampsia; however, it has not been validated in a diabetic population (34–37). This may explain why we found no difference in the S/D ratio in our study, whereas Giles et al. (20) have previously shown that a high S/D ratio correlates with significant reduction in small artery vessel count on placental histopathology.

The rate of “any neonatal complication” was elevated in both groups and can be explained by our high-risk population. Apart from SGA and fetal distress (which was of borderline significance), the presence of other adverse neonatal complications was similar between both groups. However, babies in the falling insulin requirement group were delivered slightly earlier and one in four required admission to NICU. It is possible that NICU intervention may have prevented the occurrence of other more serious complications.

In contrast to our results, three previous studies have reported no association between falling insulin requirements and adverse obstetric outcomes (6–8). The percentage of women with falling insulin requirements in these studies was lower than our cohort, however, ranging from 7.6 to 19%. The method of calculating falling insulin requirements varied in these studies and may have underestimated those at risk.

Achong et al. (6) reported weight-adjusted basal insulin requirements from 30 weeks to delivery in a retrospective cohort study of 54 women with type 1 diabetes. A ≥15% fall in insulin dose from 30 weeks’ gestation to delivery was considered significant, identifying 10% of their cohort as having falling insulin requirements. However, other studies have shown that the peak insulin dose in type 1 diabetes occurs later in the third trimester, between 32 and 37 weeks (2,8,26,38,39). In keeping with this, our study, which also included women with type 2 diabetes, showed the peak insulin dose was reached at 34 weeks (31–36) in the falling insulin requirement group and 37 weeks (35–38) in the control subjects. Calculating the change in insulin dose from 30 weeks rather than the peak dose may therefore underestimate the fall in requirements. On the other hand, calculating the fall in requirements from a later gestation increases the potential of excluding women who delivered earlier due to obstetric complications, as was the case in McManus and Ryan’s (7) 5-year retrospective cohort study where weight-adjusted insulin requirements were examined in pregnancies progressing beyond 36 weeks only. Although Steel et al. (8) did look at the percentage change in insulin from the peak to the trough dose in 236 women with type 1 diabetes, only a small proportion of women met their criteria of falling insulin requirements defined as ≥30% fall over a 7-day period. In addition, recruitment for the studies by Steel et al. (8) and McManus and Ryan (7) occurred over 20 years ago when clinical practice was vastly different. It is still unclear whether the timing and the degree of falling insulin requirements modulate adverse risk and what cutoffs are clinically relevant. Our definition of falling insulin requirements was therefore designed to increase the sensitivity of detecting women at risk for adverse outcomes while maintaining adequate power to detect significant differences. The smaller proportion of women identified as having falling insulin requirements in the previous studies could account for the conflicting results.

Whereas previous studies have focused on type 1 diabetes, this study has shown that women with type 2 diabetes are equally at risk for falling insulin requirements as those with type 1 diabetes. Type 2 diabetes is now more common than type 1 in women of childbearing age, and we have previously demonstrated that it is associated with similar or greater rates of pregnancy complications than type 1 diabetes (9). Interestingly, the small number of women with “overt diabetes in pregnancy” appeared to be at higher risk of falling insulin requirements than those with type 2 diabetes. Women with overt diabetes require prompt treatment and close follow-up to ensure glycemic targets are restored quickly (10). This may result in a rapid escalation of insulin dose, by the clinician, subsequently followed by a fall in insulin requirements once the initial period of glucotoxicity resolves. This mechanism is independent of placental function, and therefore may not translate to an increased risk of adverse obstetric outcomes. However, as the number of women with overt diabetes is small, larger prospective studies are needed to validate these findings.

We acknowledge our study has some limitations. The data were collected retrospectively; thus we were unable to quantify caloric intake, particularly changes in carbohydrate intake, which could have influenced insulin requirements. However, our results indicate a similar reduction in the basal and prandial component of the TID, making it less likely that reduced carbohydrate intake influenced the results. We were also unable to identify the timing of the fall in insulin requirements in relation to the onset of clinical outcomes and therefore cannot confirm that this preceded the adverse events. Furthermore, we were unable to discount the possibility that knowledge of falling insulin requirements resulted in additional surveillance by caregivers or if this could account for earlier delivery in this group. Falling insulin requirement was not specifically reported as the reason for intervention in any of the cases. However, if adverse outcomes were ameliorated as a consequence of any such interventions, then the true incidence of adverse outcomes associated with falling insulin requirements would be even greater than we have reported.

Another potential limitation of studies examining the calculated fall in insulin requirements is that this is determined by the peak insulin requirements, which in turn are influenced by glucose control and the vigor by which glycemic targets are pursued. However, the fact that the peak insulin requirements and HbA_1c were not different between case and control subjects suggests that this was not a significant factor in determining which subjects developed a fall in insulin requirement of ≥15% in our cohort. In keeping with the three previous studies, we used the change in exogenous insulin requirement as a proxy measure for insulin sensitivity. Ideally an objective measure of insulin sensitivity would be used; however, techniques such as the hyperinsulinemic-euglycemic clamp or homeostatic model assessment are less practical, particularly in the later stages of pregnancy, and limited in the setting of exogenous insulin treatment. As the change in insulin sensitivity may be sudden and unpredictable, the timing of such an investigation would be crucial, once again reducing the utility in this setting. The few studies that have previously examined biomarkers that could potentially be used as a measure of insulin sensitivity did not find any correlation with insulin requirements (26,40). However, these studies did not specifically examine women with falling insulin requirements. In future prospective studies, measurement of the hormones implicated in the pathophysiology of insulin resistance may be useful.

In conclusion, we have shown that falling insulin requirements in late pregnancy are associated with maternal and neonatal outcomes, suggesting placental dysfunction. These findings have important clinical implications in light of the previous negative findings and substantiate the practice of close fetal monitoring and consideration of prompt delivery in pregnancies where insulin requirements decline by >15%. Moreover, we have demonstrated a link between falling insulin requirements and preeclampsia, suggesting similarities in the underlying pathogenesis. Prospective studies are warranted to confirm these findings and to guide recommendations in clinical management.

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

Author Contributions. S.P. designed the study, collected and analyzed the data, and wrote the manuscript. M.M. contributed to critical review and edited the manuscript. N.W.C. designed the study and wrote the manuscript. S.P. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Preliminary data from this study were presented in oral form at the Australian Diabetes in Pregnancy Society Annual Scientific Meeting, Sydney, New South Wales, Australia, 30−31 August 2013, and in poster form at the 74th Scientific Sessions of the American Diabetes Association, San Francisco, CA, 13–17 June 2014.