Impact of experiencing multiple vulnerabilities on fetal growth and complications in women with hyperglycemia in pregnancy

Baseline characteristics included age, body mass index (BMI), smoking status before pregnancy, parity, history of familial diabetes, hypertension, previous HIP, macrosomia or fetal death, and sociodemographic factors. The latter included occupation, self-reported region of origin, healthcare cover, and our three socioeconomic vulnerability indicators (i.e., psychosocial deprivation, food insecurity, and French language proficiency). Participants selected their region of origin from a list of world regions provided by the study investigator. The decision to collect data on region of origin was based on observed ethnic disparities in maternal and fetal health outcomes [23].

In France, healthcare is covered by i) social security (standard health insurance, complemented by private health cover for most of the population), ii) universal health protection for low-income persons, iii) complementary universal health protection for the poorest persons (instead of private health cover), and iv) state medical aid for those who are living in France illegally. Some people have no healthcare cover, either because they do not ask for it or because they have just immigrated.

The three socioeconomic vulnerability indicators were evaluated during the first day of diabetes care. A translator was present to help women with poor proficiency in French.

First, psychosocial deprivation was measured using the 11-item French-based EPICES questionnaire. In English, EPICES stands for Psychosocial Deprivation and Health Inequalities in Health Centers. The questionnaire items focus on various aspects of socioeconomic conditions and family environment (see Additional file) [24]. For the present study, we used an EPICES score > 30.17 to define psychosocial deprivation [25].

Second, food insecurity was assessed using a single question with four possible response options as follows: “Which of these statements best describes the food eaten in your household? i) Enough of the kinds of food you want to eat; ii) Enough but not always the kinds of food you want to eat; iii) Sometimes not enough to eat; iv) Often not enough to eat”. This question is used in the U.S. Household Food Security Survey Module by the USA Department of Agriculture to measure food insecurity [10]. Participants were categorized as having food insecurity if they answered options iii) or iv).

Third, participants were asked how well they spoke French. There were three propositions: poor, moderate, or good proficiency.

The primary outcome was neonatal weight at birth according to gestational age, classified into three categories: SGA, appropriate-for-gestational-age (AGA), and LGA infant. LGA and SGA infant were defined as a birth weight greater than the 90^th percentile and lower than the 10^th percentile for the standard French population, respectively [26].

The secondary outcomes were HIP-related events, categorized into maternal and neonatal complications [27]. For maternal outcomes, we considered hypertensive disorders of pregnancy, pre-eclampsia and cesarean sections. Hypertensive disorders include hypertension before pregnancy, gestational hypertension and pre-eclampsia. Gestational hypertension was defined by the onset of hypertension (blood pressure ≥ 140 mmHg systolic or ≥ 90 mmHg diastolic) at or after 20 WG, in the absence of proteinuria and without biochemical or hematological abnormalities [28]. When earlier blood pressure values were unknown, we considered that hypertensive disorder was not present. Pre-eclampsia was defined as having a blood pressure ≥ 140/90 mmHg for two measurements four hours apart, and proteinuria of at least 300 mg/24 h or a 3 + level with dipstick testing in a random urine sample, and/or evidence of maternal acute kidney injury, liver dysfunction, neurological features, hemolysis or thrombocytopenia, and fetal growth restriction [28]. Caesarean sections were defined as selective and emergency caesarean sections before or during delivery.

For neonatal outcomes, we considered shoulder dystocia, prematurity, neonatal hypoglycemia, neonatal death and stillbirth [29]. Shoulder dystocia was defined as the use of obstetrics maneuvers. Prematurity was defined as occurring before 37 WG, and neonatal hypoglycemia as at least one blood glucose measurement under 2 mmol/L during the first two days of life. Finally, we also considered stillbirth and neonatal death (in the first 24 h of life).

For the present study, we collected data on two habits (smoking status and fruit and vegetable consumption), and on healthcare during pregnancy. During the dietary survey at HIP care initiation, all women answered the following question “Do you eat fruit and/or vegetables every day?”. Data collected on healthcare included gestational weight gain (GWG), insulin therapy (yes or no), the time between their most recent HIP outpatient visit and their first related prescription of insulin (days), the number of consultations with a diabetologist between HIP diagnosis and delivery (number), whether hospitalization was required for uncontrolled diabetes or for pregnancy complications (percentage of women requiring hospitalization, number of hospitalizations for each woman), and hospitalization duration (days). GWG was calculated as the weight just before delivery minus the self-reported pre-pregnancy weight. At Jean Verdier hospital, women receive insulin therapy (no other treatment is used in France for HIP) when pre-prandial and/or 2-h post-prandial glucose levels are greater than 5.0 and/or 6.7 mmol/L, respectively [22].

We collected all the study data with the help of other physicians, and a research study technician created the database. EC and EV ensured the accuracy of the database contents. Descriptive statistics were used to describe the entire population in terms of sociodemographic parameters, medical history, and glycemic figures. Baseline continuous variables were expressed as mean ± standard deviation. Categorical variables were expressed as frequencies (percentages).

Missing data were not replaced. With regard to French language proficiency level, we created a binary variable: poor combined ‘poor’ and ‘moderate’ proficiency, and good reflected ‘good’ proficiency.

Student’s t-test was used to compare continuous variables. The Chi-squared or Fisher’s-exact test was used for categorical variables, as appropriate. To explore the possible presence of selection bias, the baseline characteristics of the women included were compared with those not included.

Patient characteristics were compared for the three vulnerability indicators (i.e., psychosocial deprivation, food insecurity, and French language proficiency).

The three categories of neonatal weight at birth according to gestational age (i.e., primary outcome) were compared between the three secondary outcomes), as well as vulnerability indicators. We then analyzed maternal and neonatal outcomes (i.e., habits (i.e., smoking status and fruit and vegetable consumption) and care during pregnancy (i.e., explanatory variables) according to the same three indicators. To evaluate HIP-related events, we performed an exploratory analysis where we created a composite criterion that included pre-eclampsia, LGA infant, and shoulder dystocia. However, as pre-eclampsia and shoulder dystocia were rare, this analysis was not retained as it did not add any new information.

Of the potential 2,649 participants, 1,168 completed the EPICES questionnaire and answered the two questions on food insecurity and French language proficiency. These women were included in the present analysis. They were similar to non-included women except that a lower percentage were unemployed (63.4 vs. 68.4%, respectively; p = 0.01) (Additional Table).

Of those included, 31.2% had eGDM, 62.3% GDM and 6.5% DIP. Mean age was 33 years, 27% had obesity and 34.2% reported a family history of diabetes (Table 1). With regard to sociodemographic characteristics, 36.6% were employed. Almost one-fifth (19.3%) were born in Europe, 36% in North Africa, 15% Sub-Saharan Africa, 17.3% in South Asia, and 7.5% elsewhere. Just over half (55.6%) had standard health insurance. In terms of the three socioeconomic vulnerability indicators studied, 56% suffered from psychosocial deprivation, 17.9% reported food insecurity, and 27.5% reported poor French language proficiency.

Compared with women who did not suffer from psychosocial deprivation, those who did were less likely to have a family history of diabetes, to smoke, to work, and to have universal health protection and state medical aid (see Table 1). In contrast, they were more likely to be immigrants from Sub-Saharan Africa and South Asia, and more likely to have higher parity and higher FPG values from the OGTT test. Finally, women with psychosocial deprivation were more likely to report food insecurity and poor French language proficiency.

The rates of SGA, AGA and LGA infant were 11.4%, 76.5% and 12.2%, respectively (Table 2), with a similar distribution in women with and without psychosocial deprivation, and in those with and without food insecurity. Neonatal weight at birth according to gestational age was associated with French language proficiency level (Table 3). Women reporting poor proficiency were more likely to experience abnormal fetal growth (25.3% of infants overall, specifically 11.4% SGA, and 13.9% LGA) than women reporting good proficiency (18.3% overall, specifically 10.8% of SGA, and 7.5% LGA), with p = 0.02 (Table 3).

In an exploratory analysis, we explored the distribution of SGA-AGA-LGA infant according to the combination of the three indicators for socioeconomic vulnerability. A total of 880 women had available data for this analysis. Figure 2 shows that women with more than one vulnerability were more likely to have an SGA or LGA infant than those with one or no vulnerability; however, these differences were not significant.

As described in Table 2, the factors associated with LGA infant were obesity, higher parity, a history of HIP, a history of macrosomia, a higher FPG level before 22 WG, and greater gestational weight gain during pregnancy.

Women of South Asian origin were more likely to have an SGA infant (25% vs. 11.3% for SGA infant).

Multinomial logistic regression variables included BMI, parity, gestational weight gain, HIP status, macrosomia history, ethnicity, health insurance cover, French language proficiency, and newborn’s sex. Of these, parity (odds ratio (OR) 0.76 (95% interval confidence 0.59–0.99), p = 0.04), South-Asian ethnicity (OR 2.97 (1.16–7.58), p = 0.02), and poor French language proficiency (OR 0.36 (0.15–0.84), p = 0.02) were independent predictors of SGA infant. Weight gain (OR 1.04 (1.00–1.08), p = 0.045), macrosomia (OR 5.33 (2.47–11.54), p < 0.01), and complementary universal health protection (OR 2.35 (1.21–4.55), p = 0.01) were independently associated with LGA infant.

Unlike the present work, previous studies reported poorer maternal and neonatal outcomes for the most psychosocially deprived pregnant women. The inverse care law supports the notion that “the availability of good medical care tends to vary inversely with the need for it in the population served” [30]. This concept has not been studied widely in the context of pregnant women. The burden of psychosocial deprivation on pregnant women and its association with an unhealthy diet may explain the higher risk of severe pregravid obesity, higher risk of HIP and earlier diagnosis, higher gestational weight gain, lower control of glycemic figures and more frequent insulin therapy which we found in our study [4, 7, 13]. Moreover, psychosocial deprivation increases the risk of inadequate prenatal care. In one study, among 10,419 pregnancies, 23.3% of women with a low socioeconomic status either reported late prenatal care, attended fewer than 50% of planned consultations, or missed their third trimester anomaly scan [31] which screens for abnormal fetal growth [32]. Moreover, the higher the number of psychosocial vulnerabilities reported, the higher was the risk of inadequate prenatal care. Elsewhere, using a qualitative approach, Whittle et al. suggested that material need-based insecurities induce uncertainty and experiences of discrimination, leading to lower implications for health in affected women [33]. We agree with Chung’s hypothesis that a low education level may affect the pregnancy outcomes of patients with complex medical conditions [8].

We found a relationship between poor French language proficiency and abnormal fetal growth. This is not surprising given that women with a low socioeconomic status and with HIP have a higher risk of LGA [8, 34] or SGA infant [35‐37], or both [38].

Furthermore, in a previous study by our group, the risk of LGA infant was 50% higher among women with a low socioeconomic status in a cohort comprising 996 women (56% of whom had psychosocial deprivation) [34]. In China, Chung et al. demonstrated a 35% higher odds ratio of LGA infant (OR 1.35 (1.09–1.70)) in women with a middle-school level of education compared to college-educated participants [8].

In contrast, in another Chinese cohort, the most psychosocially deprived participants had a higher odds ratio of SGA infant than the least deprived (OR 1.63 (95% CI 1.29–2.08) for the 4^th quartile compared to the 1^st) [39]. In the USA, Harper et al. reported that women whose gestational weight gain was lower than the recommended value were more likely to have an SGA infant than those whose gestational weight gain was higher than that recommended (9.1% vs. 7.3%, respectively) [37]. Recently, Jardine et al. estimated that 31.1% of fetal growth restriction could be attributed to socioeconomic inequality in England; this proportion decreased to 16.4% when adjusted for ethnic group, smoking and BMI. In their cohort, women from South Asia had a higher odds ratio of fetal growth restriction [35].

Elsewhere, Wentz et al. highlighted that area-based deprivation was more strongly associated with LGA infant than with SGA infant [38]. The same authors observed variations in the level of association depending on race/ethnic group. It is difficult to reconcile all these contrasting findings.

Here we discuss some hypotheses as to why we found similar maternal and neonatal outcomes, irrespective of psychosocial deprivation or food insecurity status. First, the monocentric characteristic of our study may have reduced disparities. Second, the rate of insulin therapy was higher in the present study (43.3%) than in a previous study by our team on psychosocial deprivation in women with HIP from the same Paris suburb (29.4%) [7]. It was also higher than in a Japanese cohort (8.6% of women with HIP (defined according to IADPSG/WHO criteria) where the mean BMI was 22.2 kg/m² [40]). In other studies, this information was not reported [36, 37]. Third, most women in our sample had psychosocial deprivation, and healthcare providers are used to treating this population. Fourth, the mean number of pregnancy consultations was also relatively high, and higher in women with food insecurity and/or poor French language proficiency. This may have translated into better tailoring of insulin therapy to the individual’s needs, and therefore better pregnancy outcomes. Increasing inter-professional communication between the specialists in our hospital (obstetricians, dieticians and diabetologists) could lead to more frequent and effective follow-up of diabetes. Indeed, glycemic monitoring is not performed only by diabetologists; it involves the participation of various specialists.

We observed a higher rate of SGA infant in women with poor French language proficiency. A previous study in Canada highlighted the consequences of language barriers on pregnancy outcomes. Specifically, it found that non-English-speaking Asian women had a lower risk of LGA infant than English-speaking women [16]. We hypothesize that women with poor French language proficiency in our sample may have been more afraid of their diagnosis of HIP, as they would not have been able to understand the complex related vocabulary and arguments (controlled carbohydrate diet, fruit snacks, six daily glucose control tests, different fasting and postprandial goals, weight follow-up, insulin requirements, etc.), and that the consequence of this may have been an overly restrictive diet in order to avoid gaining weight [41]. In our study, the mean gestational weight gain was 8.2 ± 4.4 kg vs. 9.1 ± 5.6 kg, respectively, for women with good proficiency.

However, in the multivariable analysis (data not presented), the main factor associated with SGA infant was South Asian ethnicity, while poor proficiency was an independent protective factor. This finding mitigates our initial hypothesis that women restrict their consumption of certain food types. Furthermore, Venkatesh et al. published data on maternal and fetal outcomes for women with HIP in the USA over a period of six years. They reported a higher risk of SGA infant in persons of Asian/Pacific island origin than in white women (OR 1.84 (1.82–1.87)), after multiple adjustment [42].

There was a relatively high rate of pre-eclampsia (3.3% (range 1.3–5%)) in our study population, especially in women with food insecurity (5%). The high SGA infant rate in this group (13.2%) would suggest overrepresentation of fetal growth restriction, due to a possible combination of endogenic and exogenic oxidative stress with placental dysfunction, reflected in the higher rate of pre-eclampsia observed [43]. Women with psychosocial deprivation or food insecurity reported an unhealthy diet more often than those without these vulnerabilities. The consumption of fruit, vegetables, cereals and grains is necessary for fetal growth because of their selenium content (on average, 41% of total selenium intake); other sources of selenium are fish and seafood (29%), meat (23%), and egg and dairy products (20%) [44].

The strengths of our study include a large study sample with three socioeconomic vulnerability indicators to evaluate socioeconomic deprivation (French language proficiency level being investigated for the first time), a multi-ethnic cohort with adjustments being made for confounders, and similar prenatal care provided by specialists, diabetologists and obstetricians to all those included.

The study also has several limitations. First, as only women who provided complete data (i.e., fully completed the EPICES questionnaire, answered the question on food insecurity, and the question on French language proficiency) were included, they differed somewhat from those not included (Additional Table). Second, as French language proficiency was self-reported instead of using a reading/writing/listening test, social desirability bias cannot be excluded. Third, the EPICES psychosocial indicator was initially validated using only people from the French ethnic group. In our study, this group accounted for less than 20% of the study population. Finally, confounding bias may have been induced by the association between ethnicity and fetal growth. Most women with poor language proficiency were from south Asia; this ethnicity has already been associated with a higher risk of SGA infant [42].