Association of Parity With Insulin Resistance Early in Pregnant Women: ECLIPSES Study

Abstract

Context

Little is known about whether parity is associated with elevated early-pregnancy insulin resistance (IR), or whether overweight/obesity contributes to increasing the possible effect.

Objective

We determined the associations between parity and glucose metabolism parameters in the first trimester of pregnancy in a Mediterranean pregnant population, and whether these associations are affected by overweight/obesity.

Methods

A cross-sectional study was conducted of 264 healthy pregnant women from the ECLIPSES study who were recruited at 12 weeks of gestation. At baseline, details on socioeconomic status, obstetric history (including parity, ie, number of births), lifestyle factors, anthropometry, and blood samples were collected. Fasting serum glucose, insulin, and homeostasis model assessment of insulin resistance (HOMA-IR) index were assessed in the first trimester. Elevated IR was defined as the upper HOMA-IR tertile (≥1.58). Multivariable linear regression and Cox regression model with constant time were performed.

Results

Parity ranged from 0 to 4. After multivariable adjustment, the insulin levels (β [% change]: 20.92; 95% CI, 4.08-37.71) and HOMA-IR index (β [% change]: 19.72; 95% CI, 2.43-40.49) were positively associated with parity. Additionally, multiparous women, as compared to nulliparous, were more likely to have higher HOMA-IR levels (primiparous [1 birth], β [% change[: 16.88; 95% CI, −1.00 to 37.99; multiparous [≥2 births), β [% change]: 32.18; 95% CI, 3.56-68.71), and an increased relative risk (RR) of an elevated IR (primiparous [1 birth], RR: 1.55; 95% CI, 1.03-2.36; multiparous (≥2 births), RR: 1.72; 95% CI, 1.05-2.83). The combination of multiparity and overweight/obesity conferred a 3.04-fold increase in the RR of elevated IR, which increased proportionally to the number of parities.

Conclusion

This study demonstrates that parity may have a negative effect on early-pregnancy IR and that maternal overweight/obesity appears to further aggravate this relationship.

parity, multiparous, insulin resistance, pregnancy, ECLIPSES

Pregnancy is a time-limited condition, in which women undergo significant physiological and metabolic changes. These changes lead to maternal fat accumulation and a gradual increase in peripheral insulin resistance (IR) (1) throughout pregnancy to accommodate the growing fetus (2). However, it is important to note that excessive maternal hyperglycemia- and hyperinsulinemia-induced IR in early pregnancy is a hallmark of a metabolic disorder. This condition has recently emerged as the prominent cause of pregnancy complications and adverse perinatal outcomes, including gestational diabetes mellitus (GD) (3); preeclampsia (4); macrosomia, being large for gestational age; and cesarean delivery (5).

Moreover, although earlier studies (1, 2, 6, 7) suggest that maternal IR returns to prepregnancy levels by 1 year postpartum, the recent literature casts some doubt on this premise (8, 9). It has been hypothesized that recurrent maternal IR episodes due to repeated pregnancies lead to a progressive worsening of the glucose tolerance in each pregnancy, manifesting in GD or it may even permanently disturb glucose homeostasis in women in later life (8-11). Nevertheless, the underlying mechanisms for parity-related IR during pregnancy are largely unknown, complex, and most likely reflect a range of factors, such as placental hormones, lifestyle modifications, and genetic and epigenetic contributions (12).

In this regard, the effect of the multiparity on the GD or recurrent GD during later pregnancies has been a topic of research for the last few years (11, 13). Nonetheless, the results are not completely convincing, that is, the high rate of GD among multiparous women could be due to the confounding effect of higher maternal age or body adiposity (14, 15). In addition, the diagnosis of GD is not made until the latter half of pregnancy (at 24-28 weeks) (16), which might be too late to completely reverse the intrauterine hyperglycemia-induced adverse effects on offspring that can occur in the early stages of pregnancy (17). Alternatively, maternal IR in the first trimester, assessed by the homeostasis model assessment of the IR index (HOMA-IR), has been proposed as a reliable marker to predict subsequent GD (3, 18). In this context, there are few studies examining whether repeated parity could be associated with an increased risk of IR in very early pregnancy. This is possibly because IR is not routinely assessed during prenatal examinations.

To the best of our knowledge, only 3 studies have researched this relationship; however, their findings are inconsistent (19-21). Two studies have found a positive association between IR in middle pregnancy (20-30 weeks’ gestation) and parity (20, 21), which was not confirmed by the study by Seghieri et al (19). In this last study, the authors (19) found that parity is not directly linked to insulin sensitivity/secretion during the last trimester in pregnant women at high risk of GD; however, high pregestational body mass index (BMI) was strongly related to a considerable impairment in insulin sensitivity during this period. In accordance with these findings, at least in part, we recently found that maternal BMI in early pregnancy is more relevant than gestational weight gain for predicting cardiometabolic risk during pregnancy (22).

Nevertheless, it remains undetermined whether parity confers an independent effect on early-pregnancy IR, and particularly in a healthy Mediterranean pregnant population, in which the sociodemographic and Mediterranean lifestyle traits of women can be regarded as protective factors against IR. It is also necessary to determine whether parity contributes, in combination with high maternal BMI, to a worse IR. Thus, the aim of the present study is to assess how parity is related to glucose metabolism parameters measured early in pregnancy in a pregnant population from a Mediterranean region of northern Spain, and assess whether these associations vary according to overweight/obesity status.

Materials and Methods

Study Design and Participants

We conducted a retrospective study analyzing data from 264 healthy pregnant women with singleton pregnancies and without previous history of diabetes. These women had data available on parity history, fasting serum glucose, and insulin concentrations in the first trimester (at ∼12 weeks’ gestation). Women participated in the ECLIPSES (Ensayo CLInico Para Suplementar con hierro a EmbarazadaS) study, in which a total of 791 pregnant women were recruited during their first antenatal visit in 12 sexual and reproductive health care services (ASSIR) of the Catalan Institute of Health (Catalonia, Spain) between 2013 and 2017. Briefly, the ECLIPSES aimed to determine the highest level of effectiveness of iron supplementation based on hemoglobin levels in early pregnancy to optimize maternal and child health. Details of the study's protocol, as well as inclusion/exclusion criteria, have been described elsewhere (23). The ECLIPSES study was registered at www.clinicaltrialsregister.eu (ID: EUCTR-2012-005480-28) and at www.clinicaltrials.gov (ID: NCT03196882). Ethical approval for the study was obtained from the Jordi Gol Institute for Primary Care Research and the Pere Virgili Institute for Health Research. The research complies with the tenets of the Helsinki Declaration. Participants provided written informed consent.

Data Collection

Midwives and nutritionists collected data on demographic characteristics (age, socioeconomic information, and education level), health behaviors (physical activity [PA], smoking, and diet), medical and obstetric history (planned pregnancy [yes, no]), as well as anthropometric measurements in the first trimester of pregnancy (at week 12). Familiar socioeconomic status (SES) was calculated by combining information on occupational status, classified according to the Catalan classification of occupations (CCO-2011), and educational level. It was then classified as low, middle, or high. The women's educational level was classified into 3 groups: low (primary school or less), medium (secondary studies), and high (university studies or more).

PA was measured using the short version of the International Physical Activity Questionnaire (IPAQ-S) (24). This was derived from total metabolic equivalents (METs-min/week) values based on frequency and duration of walking and moderate and vigorous-intensity activity and divided into tertiles (T1: <1070, T2: 1070-3336, T3: ≥ 3336 METs-min/week).

Smoking was assessed with the Fagerström questionnaire (25) and women were classified into 3 groups: current, former, and never smokers. Eating habits were assessed by dietitians using a 45-item self-administered food frequency questionnaire validated in our population (26). Herein, we focused on women's overall diet quality assessed by using a relative Mediterranean Diet (rMedDiet) score, which has been used in the previous ECLIPSES studies (27). For this study, based on the participant distributions, the continuous rMedDiet score (ranging from 0 to 18 points) was categorized into tertiles (T1: <9, T2: 9-12, T3:≥12 points). Alcohol consumption was assessed as yes or no.

Maternal weight (kg) and height (cm) were also measured. Early-pregnancy BMI (ep-BMI) (ie, at week 12) was calculated from these measurements (weight [kg]/height [m²]), and women were classified into 2 groups: normal weight (NW, ep-BMI <25.0) and overweight/obesity (OWO, ep-BMI ≥25.0) for this analysis.

Parity Assessment

Parity was defined as the number of singleton pregnancies of at least 20 weeks (regardless of whether the child was liveborn) as reported by women on the interviewer-administered questionnaire. For this analysis, pregnant women were classified as nulliparous (no prior viable pregnancies) or multiparous (given birth to 1-4). To more accurately examine the effect of parity, in secondary analyses women were categorized as having no children (nulliparous), 1 child, and 2 or more children.

Outcome Ascertainment

Blood samples were collected from pregnant women in the first trimester of pregnancy (12 weeks). Serum was separated from blood cells by centrifugation and stored in Biobank at −80 °C until analysis of the fasting glucose and insulin. We included in the analysis only women who underwent blood tests after an overnight fast. Fasting glucose concentrations were measured using standard automated enzymatic methods. The coefficient of variation was 1.74%. Fasting insulin level was measured by chemiluminescence immunoassay on an ADVIA Centaur analyzer using a commercial kit (ADVIA Centaur IRI, Siemens Healthcare Diagnostics Inc). The lower and upper limits were 0.5 and 300 mUI/L, respectively, and coefficient of variation was 4.88%. All measurements were conducted at the Institut Català de la Salut Camp de Tarragona-Terres de l’Ebre–accredited laboratory, Joan XXIII University Hospital in Tarragona.

IR was assessed according to the HOMA-IR index, calculated according to the following equation: HOMA-IR = [fasting glucose (mmol/L) × fasting insulin (μIU/mL)]/22.5. The HOMA-IR index was analyzed as a continuous variable, with a larger HOMA-IR value indicating more severe IR. In addition, to classify the group at higher risk, HOMA-IR was categorized in tertiles (cutoff points were 1.04 and 1.58), and, while considering statistical power, IR was defined as being in the upper tertile. This HOMA-IR threshold of 1.58 is within the range of the cutoff values of HOMA-IR (1.51-2.31) in the first trimester for predicting GD (18).

Statistical Analysis

Data were analyzed using STATA (15.0, Stata Corp LP). Descriptive data are expressed as mean ± SD for quantitative variables and number (%) for categorical variables. Between-group comparisons were performed with t test, chi-square, or analysis of variance test, as appropriate. Since insulin and HOMA-IR were right-skewed, as expected, both were log-transformed to improve normality prior to analysis.

Unadjusted and multivariable-adjusted linear regression models were fitted to estimate the associations of parity, as the main exposure variable, separately with each outcome (fasting glucose, insulin, and HOMA-IR levels). Parity was analyzed using 2 different approaches: as a binary categorical explanatory variable (nulliparous: reference and multiparous), and as an ordinal categorical explanatory variable with 3 categories (0 [nulliparous[: reference, 1 child, and >2 previous children). An additional separate linear regression analysis was also performed to evaluate the joint association of parity and ep-BMI–based weight status in 2 groups (NW and OWO) as predictor, with each outcome. For this analysis, women were grouped into 4 categories: nulliparous + NW (reference), multiparous + NW, nulliparous + OWO, and multiparous + OWO. Based on previously known or association/risk factors for either the exposure, the outcome, or both (12, 22, 28), we considered the following a priori selected covariates as possible confounders: age (<25: reference, 25-29, ≥ 30 years), ep-BMI categories (NW: reference, OWO; except in the parity + ep-BMI analysis, where it was integrated into the composite explanatory variable itself), educational level (low [primary school or less]/medium [secondary studies]: reference, high [university studies or more]), smoking status (never smoker: reference, current/former smoker), alcohol consumption (no: reference, yes), planned pregnancy (no: reference, yes), PA (as T: T1 ≤ 1070: reference, T2 1071-3335, T3 ≥ 3336 METs-min/week), and rMedDiet score (as T: T1 ≤ 8: reference, T2 9-11, T3 ≥ 12 points). Because dietary covariates (MedDiet score and alcohol intake, both 3.2%, n = 9) had missing values, we employed multiple imputation with the chained-equations method to impute missing data (29) based on the correlation of missing variables with other participant characteristics such as maternal age and BMI. For each analysis, we created 20 imputed data sets and pooled the results using the “mi” command in Stata. Estimates were presented as β coefficients (β) with 95% CIs. In the case of log-transformed outcomes (insulin and HOMA-IR), estimates were reported as percentage changes calculated using the equation: (e^β–1) × 100%). The multicollinearity test was carried out by looking at the tolerance (1/VIF) values and variance inflation factors (VIFs). All tolerance values were greater than 0.10 and all VIFs were less than 2.0, so there was no multicollinearity.

In addition, a separate multivariable-adjusted Cox regression model with constant follow-up time set at t = 1 for all individuals (given the cross-sectional design) and robust variance rather than logistic regression was applied to estimate the relative risks (RR) and 95% CI (30) for elevated HOMA-IR index (≥1.58 points) according to parity (nulliparous as reference) and parity + ep-BMI categories (nulliparous + NW as reference) (in separate models); these included the same covariates as the linear models. A test for linear trend was calculated by treating ordinal categorical exposure variable as continuous variable.

Finally, supplementary subgroup multivariable analyses, examining the association of parity with each outcome, were performed stratified by ep-BMI–based weight status (NW and OWO). The interaction between weight status and parity in these associations was assessed by calculating the likelihood ratio test between the fully adjusted model and the same model, including the interaction product term, in the complete data set (n = 255), as it cannot be performed with multiply imputed data. The level of statistical significance was set to 2-sided P values less than .05.

Results

The characteristics of the study participants are shown in Table 1. The average age of our sample was 29.6 ± 4.7 years, and a large percentage of the women were older than 30 (57%). The mean ep-BMI was 24.1 ± 3.5, which falls within the normal classification (18.5-24.9), but the prevalence of overweight/obesity was 36.0% at early pregnancy. The analyzed cohort of women did not exhibit significant differences in sociodemographic and lifestyle characteristics when compared to nonparticipants, except for age and ep-BMI, which was lower (both P < .05) (data not shown). The parity of our population ranged from 0 to 4; 57.6% of women were multiparous women. The multiparous women were significantly older, less educated, and more likely to be from a higher SES, than nulliparae (see Table 1).

Table 1.

Sociodemographic and lifestyle characteristics of pregnant women

General characteristics	Overall (n = 264)	Parity		Pa
		Nulliparous women (n = 112)	Multiparous women (n = 152)
Age, mean ± SD, y	29.6 ± 4.7	28.6 ± 4.5	30.6 ± 4.7	<.001c
Age categories, n (%), y
<25	40 (15)	25 (22)	15 (10)
25-29	73 (28)	34 (30)	39 (26)	.005b
≥30	152 (57)	53 (47)	98 (64)
Weight, mean ± SD, kg	63.3 ± 9.6	62.9 ± 9.9	63.5 ± 9.5	.57
Ep-BMI, mean ± SD	24.1 ± 3.5	23.9 ± 3.7	24.3 ± 3.4	.48
Ep-BMI categories, n (%)
Normal weight (BMI < 25)	169 (64)	76 (68)	93 (61)
Overweight/obesity (BMI ≥ 25)	95 (36)	36 (32)	59 (39)	.264
Educational level, n (%)
Low (≤ primary)	83 (31)	28 (25)	55 (36)
Medium (secondary)	97 (37)	36 (32)	61 (40)	.004b
High (≥ university)	84 (32)	48 (43)	36 (24)
Familiar SES, n (%)
Low	35 (13)	7 (6)	28 (18)
Medium	180 (68)	77 (69)	103 (68)	.003b
High	49 (19)	28 (25)	21 (14)
Smoking status, n (%)
Never smoker	185 (70)	74 (66)	111 (73)
Former smoker	42 (16)	23 (20)	19 (13)	.21
Current smoker	37 (14)	15 (13)	22 (14)
Alcohol consumption
No	222 (84)	95 (85)	127 (84)	.30
Yes	33 (13)	11 (10)	22 (14)
Missing	9 (3)	6 (5)	3 (2)
Physical activity (METs-min/wk)
T1 (≤1070)	87 (33)	36 (32)	51 (34)
T2 (1071-3335)	117 (44)	51 (45)	66 (43)	.94
T3 (≥3336)	60 (23)	25 (22)	35 (23)
rMedDiet score (point)
T1 (≤8)	92 (35)	37 (33)	55 (36)
T2 (9-11)	107 (41)	41 (37)	66 (44)	.34
T3 (≥12)	56 (21)	28 (25)	28 (18)
Missing	9 (3)	6 (5)	3 (2)
Planned pregnancy
No	62 (23)	22 (20)	40 (26)	.21
Yes	202 (77)	90 (80)	112 (74)

Values are expressed in means ± SD or number (%, percentage).

Abbreviations: Ep-BMI, early-pregnancy body mass index; MET, metabolic equivalent; rMedDiet, Mediterranean diet; SES, socioeconomic status; T, tertile.

^aP value for the differences across primiparous vs multiparous women as derived from t test or chi-square test, as appropriate.

^bThe significance of the numbers in bold is P less than .05 compared with reference category.

^cThe significance of the numbers in bold is P less than .001 compared with reference category.

The average fasting glucose, insulin, and HOMA-IR levels were 70.2 ± 10.7 mg/dL, 7.77 ± 1.76 mU/L, and 1.32 ± 1.85 points, respectively, for the total study population (Table 2). We observed that multiparous women had higher insulin and HOMA-IR levels than nulliparous women. Similarly, the HOMA-IR index was significantly higher in overweight/obese women compared to NW peers. This gradually increased according to parity and ep-BMI categories, and the mean HOMA-IR was significantly higher in the multiparous + OWO group compared with the nulliparous + NW group (see Table 2). The fasting glucose concentrations were similar in all groups.

Table 2.

Comparison of glucose metabolism parameters by parity, maternal early-pregnancy body mass index (ep-BMI)–based weight status, and combined parity + ep-BMI–based weight status categories in the first trimester of pregnancy

Variables	No.	Glucose, mg/dL	Insulin, mU/La	HOMA-IRa
		Mean ± SD	Mean ± SD	Mean ± SD
Overall	264	70.2 ± 10.7	7.77 ± 1.76	1.32 ± 1.85
Parity
Nulliparous	112	69.8 ± 9.4	6.89 ± 1.75	1.18 ± 1.82
Multiparous	152	70.4 ± 11.5	8.41 ± 1.78c	1.44 ± 1.87c
Ep-BMI categories
Normal weight (BMI < 25)	169	69.5 ± 10.4	6.89 ± 1.70	1.17 ± 1.84
Overweight/obesity (BMI ≥ 25)	95	71.5 ± 11.4	9.49 ± 1.70c	1.65 ± 1.79c
Parity + ep-BMI categories
Nulliparous + NW	76	68.5 ± 9.3	6.35 ± 1.71	1.07 ± 1.78
Multiparous + NW	93	70.4 ± 11.3	7.39 ± 1.78	1.26 ± 1.88
Nulliparous + OWO	36	72.8 ± 9.4	8.16 ± 1.77	1.46 ± 1.84
Multiparous + OWO	59	70.7 ± 12.2	10.48 ± 1.65c	1.80 ± 1.75c

Values are expressed in means ± SD.

Abbreviations: ep-BMI, early-pregnancy body mass index; HOMA-IR, homeostasis model assessment of insulin resistance; NW, normal weight (BMI < 25); OWO, overweight/obesity (BMI ≥ 25).

^aGeometric means of log-transformed.

Statistical significance of numbers in bold is ^bP less than .05 and ^cP less than .001 compared with the first category as derived from t test or analysis of variance, as appropriate.

We further tested for possible associations between parity and each glucose metabolism parameter separately with a univariate- and multivariable-adjusted linear regression model (Table 3). In the univariate model, multiparity was significantly associated with higher fasting insulin levels (β coefficient for % change: 22.14; 95% CI, 6.72-40.49; P = .006) and a higher HOMA-IR index (β [% change]: 23.37; 95% CI, 6.18-43.33; P = .011). These variables increased significantly from 1 to 2 or more parities (P-for-trend >.007). In the fully adjusted main effects model, these associations did not change appreciably from the unadjusted associations. Multiparity was consistently related to the insulin levels (β [% change]: 20.92; 95% CI, 4.08-37.71; P = .016) and the HOMA-IR index (β [% change]: 19.72; 95% CI, 2.43-40.49; P = .03) independently of confounding factors such as maternal ep-BMI modeled either categorically or in its continuous form (data not shown). We observed similar positive trends between parity and insulin and the HOMA-IR index in OWO and non-OWO women (Supplementary Table S1 (31)), although the association of parity or 2 or more with the HOMA-IR index, compared to being in the nulliparous group, seemed to be stronger in OWO women. However, cross-product terms between weight status and parity for their associations with each outcome (glucose, insulin, and HOMA-IR) were not statistically significant (P > .14 for all interactions, as shown in Supplementary Table S1 (31)). When the parity number was increased in combination with being OWO, that is, when analyzed together, the positive relationships with insulin and HOMA-IR gradually and significantly increased (see Table 3). Specifically, in the full model, the group of multiparous + OWO women was associated with a greater increase in mean insulin levels of 64.87% (95% CI, 34.99-101.38; P < .001) and in the mean HOMA-IR index of 68.20% (95% CI, 35.12-110.01; P < .001) compared with those in the nulliparous + NW group (see Table 3). The multivariable-adjusted geometric means of the HOMA-IR according to combined parity + maternal ep-BMI categories are shown in Fig. 1.

Figure 1.

Multivariable-adjusted geometric means and 95% CIs of HOMA-IR index in the first trimester of pregnancy according to parity and maternal early-pregnancy BMI categories (n = 264). Adjusted means were calculated by analysis of covariance (ANCOVA). Models were mutually adjusted for age categories (<25 [ref.], 25-29, ≥ 30 years), educational level (low/medium [ref.], high), smoking status (never smoker [ref.], current/former smoker), alcohol consumption (no [ref.], yes), planned pregnancy (no [ref.], yes), physical activity tertile (T1: ≤ 1070 [ref.], T2: 1071-3335, T3: ≥ 3336 METs-min/week), and Mediterranean diet score tertile (T1: ≤ 8 [ref.], T2: 9-11, T3: ≥ 12 points). *P less than .05 and **P less than .001 compared with the reference category (nulliparous + NW). Points represent adjusted geometric means and whisker plots represent 95% CIs. Horizontal line (dash line) represent the overall geometric mean. BMI, body mass index; HOMA-IR, homeostasis model assessment of insulin resistance; MET, metabolic equivalent; NW, normal weight (BMI < 25); OWO, overweight/obesity (BMI ≥ 25).

Table 3.

Linear regression models of the associations of parity (modeled separately as nulliparous vs multiparous and as parity categories [0: nulliparous, 1, and ≥2 children]) and combined parity + early-pregnancy body mass index categories with glucose metabolism parameters

	No.	Glucose, mg/dL		Insulin, mU/L (% change)a		HOMA-IR index (% change)a
		Model 1 (n = 264) β (95% CI)	Model 2 (n = 264) β (95% CI)	Model 1 (n = 264) β (95% CI)	Model 2 (n = 264) β (95% CI)	Model 1 (n = 264) β (95% CI)	Model 2 (n = 264) β (95% CI)
Parity
Nulliparous	112	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.
Multiparous	152	0.60 (−2.03 to 3.24)	0.15 (−2.68 to 3.00)	22.14 (6.72 to 40.49)c	20.92 (4.08 to 37.71)c	23.37 (6.18 to 43.33)c	19.72 (2.43 to 40.49)c
Nulliparous	112	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.
1 previous child	115	0.35 (−2.46 to 3.16)	−0.24 (−3.21 to 2.72)	19.72 (3.67 to 39.10)c	18.06 (1.41 to 37.44)c	19.94 (2.12 to 40.78)c	16.88 (−1.00 to 37.99)
≥2 previous children	37	1.38 (−2.63 to 5.40)	1.71 (−2.67 to 6.11)	30.47 (5.65 to 61.28)c	29.18 (3.28 to 61.61)c	32.45 (5.34 to 66.70)c	32.18 (3.56 to 68.71)c
P for trendb		.52	.58	.007	.010	.011	.014
Parity + ep-BMI categories
Nulliparous + NW	76	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.
Multiparous + NW	93	1.87 (−1.39 to 5.12)	1.53 (−1.94 to 5.01)	15.03 (−2.47 to 35.93)	15.49 (−2.96 to 37.71)	17.47 (−1.98 to 40.49)	17.35 (−3.25 to 42.62)
Nulliparous + OWO	36	4.31 (0.04 to 8.57)c	3.95 (−.39, 8.28)	28.27 (3.15 to 59.36)c	27.51 (2.12 to 59.20)c	36.48 (7.79 to 72.98)c	34.99 (6.18 to 72.29)c
Multiparous + OWO	59	2.18 (−1.48 to 5.83)	1.64 (−2.29 to 5.58)	66.70 (38.40 to 99.37)d	64.87 (34.99 to 101.38)d	69.89 (39.10 to 109.59)d	68.20 (35.12 to 110.01)d
P for trendb		.152	.265	<.001	<.001	<.001	<.001
Parity + ep-BMI categories
Nulliparous + NW	76	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.
1 previous child + NW	71	2.09 (−1.39 to 5.57)	1.57 (−2.07 to 5.20)	13.88 (−4.88 to 34.99)	15.03 (−4.88 to 39.10)	16.18 (−4.30 to 40.49)	16.18 (−4.88 to 42.76)
≥2 previous children + NW	22	1.13 (−3.96 to 6.24)	1.56 (−4.02 to 7.15)	20.92 (−6.76 to 56.83)	18.53 (−10.42 to 58.41)	22.14 (−8.61 to 61.61)	22.14 (−10.42 to 66.53)
Nulliparous + OWO	36	4.30 (0.04 to 8.57)c	3.97 (−0.36 to 8.31)	28.40 (3.05 to 60.00)c	28.40 (2.22 to 60.01)c	36.34 (7.25 to 73.33)c	35.53 (6.18 to 71.60)c
1 previous child + OWO	44	1.15 (−2.83 to 5.15)	.37 (−3.91 to 4.64)	63.23 (32.31 to 99.37)d	58.41 (27.12 to 99.37)d	63.23 (31.00 to 105.44)d	59.36 (24.61 to 101.38)d
≥2 previous children OWO	15	5.16 (−0.79 to 11.13)	5.44 (−0.86 to 11.45)	78.60 (31.00 to 141.09)d	85.89 (33.64 to 156.00)d	91.55 (37.71 to 166.45)d	99.37 (40.48 to 185.38)d
P for trendb		.129	.192	<.001	<.001	<.001	<.001

Linear regression models were used to calculate the β coefficient (β) and 95% CI. Model 1: unadjusted model. Model 2: model adjusted for age categories (<25 [ref.], 25-29, ≥ 30 years), BMI categories (NW [ref.], OWO; not for adjusted in the parity + ep-BMI categories analysis, where it was integrated into the composite explanatory variable itself), educational level (low/medium [ref.], high), smoking status (never smoker [ref.], current/former smoker), alcohol consumption (no [ref.], yes), planned pregnancy (no [ref.], yes), physical activity tertile (T1: ≤ 1070 [ref.], T2:1071-3335, T3: ≥ 3336 METs-min/wk), and Mediterranean diet score tertile (T1: ≤ 8 [ref.], T2: 9-11, T3: ≥ 12 points).

Abbreviations: ep-BMI, early-pregnancy body mass index; HOMA-IR, homeostasis model assessment of insulin resistance; MET, metabolic equivalent; NW, normal weight (BMI < 25); OWO, overweight/obesity (BMI ≥ 25); Ref., reference.

^aNatural log-transformed; parameter estimates (β coefficient) have been back transformed to reflect the percentage change in each outcome (insulin and HOMA-IR) associated with a 1-unit increase in each exposure variable.

^bP value for trend was calculated by treating ordinal categorical exposure variable as continuous variable

Statistical significance of numbers in bold is ^cP less than .05 and ^dP less than .001 compared with the reference category.

According to the HOMA-IR cutoff point of 1.58 or greater, the prevalence of elevated IR was 41.0% and 49.0% for multiparous women and OWO women, respectively. Among the multiparous and OWO women, the prevalence increased to 57.6% (Fig. 2). After adjustment for confounders, multiparity was associated with a 1.59-fold (95% CI, 1.07-2.36; P = .021) increase in the relative risk of elevated IR and, with a trend (P-for-trend = .018) toward increased parity-associated risk of elevated IR as parity increased (parity = 1, RR: 1.55; 95% CI, 1.03-2.36; P = .036 and parity ≥ 2, RR: 1.72; 95% CI, 1.05-2.83; P = .031). Interestingly, multiparous women with OWO were nearly 2 times as likely to have elevated IR than the total multiparous population (RR: 3.04; 95% CI, 1.70-5.47; P < .001). The biggest difference in the RR was between the parity of 2 or more children + OWO group and the nulliparous + NW group (RR: 3.85; 95% CI, 1.98-7.76; P < .001). Results are shown in detail in Fig. 2.

Figure 2.

Multivariable-adjusted RR (95% CI) for elevated HOMA-IR according to parity and combined parity + ep-BMI categories. Multivariable-adjusted Cox regression models with constant follow-up time set at t = 1 for all individuals and robust variance were used to calculate the RR and 95% CI. †Models were mutually adjusted for age categories (<25 [ref.], 25-29, ≥ 30 years), BMI categories (NW [ref.], OWO; not for adjusted in the parity + ep-BMI categories analysis, where it was integrated into the composite explanatory variable itself), educational level (low/medium [ref.], high), smoking status (never smoker [ref.], current/former smoker), alcohol consumption (no [ref.], yes), planned pregnancy (no [ref.], yes), physical activity tertile (T1: ≤ 1070 [ref.], T2: 1071-3335, T3: ≥ 3336 METs-min/week), and Mediterranean diet score tertile (T1: ≤ 8 [ref.], T2: 9-11, T3: ≥ 12 points). Statistical significance of numbers in bold was set at P less than .05. Diamonds represent RR and whisker plots represent 95% CIs. BMI, body mass index; ep-BMI, early-pregnancy BMI; HOMA-IR, homeostasis model assessment of insulin resistance; NW, normal weight (BMI < 25); OWO, overweight/obesity (BMI ≥ 25); Ref., reference; RR, relative risk. *Top tertile of HOMA-IR (≥ 1.58). P for trend was calculated by treating ordinal categorical exposure variable as continuous variable.

Discussion

This study researched the relationship between parity and glucose metabolism parameters in the first trimester of pregnancy among healthy pregnant women from a Mediterranean region. We found a statistically significant positive association between multiparity and IR as assessed by HOMA-IR index, even after considering several traditional confounding factors. Women with one or more parity also had a 1.59-fold increased RR of having elevated IR compared with nulliparous women, which increased proportionally to the number of parities. This positive relationship between parity and greater RR of IR was more robust in combination with being OWO.

The results of our study highlight the clinical relevance of early assessment for IR in pregnant women who have had multiple pregnancies, particularly in pregnant women with OWO. According to our selected HOMA-IR threshold of 1.58, the prevalence rate of early-pregnancy IR among multiparous pregnant women was 41.0%, about 50% among the OWO group, and 57.6% among women who met both conditions. Although there is no standardized cutoff value of HOMA-IR for identifying IR in pregnancy (32), it is worth noting that Cohen et al (33) validated the use of HOMA-IR as a surrogate for the hyperinsulinemic-euglycemic clamp technique “gold standard” as a measure of IR in early pregnancy, even in obese women. In our study, the decision to define the upper tertile of HOMA-IR as the threshold for IR could be viewed as somewhat arbitrary; however, it is not inconsistent with published data on other measures of IR that identify IR individuals as being in the top tertile (34, 35). Furthermore, maternal IR in early pregnancy, at HOMA-IR cutoff levels within the range from 1.51 to 2.31, has been reported as a reliable marker of subsequent GD (18, 36). Our HOMA-IR threshold of 1.58 is well within this range. In this context, previous researchers have also indicated that women with high HOMA-IR in early to mid-pregnancy have an increased risk of subsequent preeclampsia, excessive weight gain during pregnancy, and of giving birth to macrosomic and large-for-gestational-age neonates (4, 5, 37). Thus, exposure in early pregnancy to elevated maternal IR, especially in multiparous women, should not be ignored.

Regarding glucose homeostasis disorders in pregnancy related to repeat pregnancies, extensive research has been carried out to investigate the relationship between multiparity and GD or recurrent GD during later pregnancies (11, 13). However, few studies have focused on maternal IR in early pregnancy in healthy pregnant women without previous history of diabetes.

Supporting our results in the first trimester of pregnancy (∼12 weeks), an earlier study conducted by Abdelsalam and Elamin (20) involving 300 pregnant Sudanese women aged 17 to 35 years in Khartoum State (Sudan) reported that both maternal fasting insulin levels and HOMA-IR in middle pregnancy (∼20-30 weeks’ gestation) increased significantly with higher parity compared to nulliparity. The highest levels were observed in grand multiparity (>5 times). Similarly, Jinlan et al (21), analyzing the data of 208 Chinese pregnant women aged 25 to 35 years in the Huzhou region, Zhejiang Province (Southeast China), stated that women with a second pregnancy were more likely to have higher glucose intolerance and HOMA-IR–based IR during the second to third trimester compared to primiparas (first-time pregnancy). In another study conducted by Seghieri et al (19) in a selected group of 1880 third-trimester pregnant women at higher risk for GD (ie, all women had glucose intolerance, OWO, and advanced age, >29 years), in the Pistoia area (Tuscany, Italy), it was found that ISI_OGTT-based insulin sensitivity during the third trimester decreased significantly only in pregnant women with parity greater than 3 compared to nulliparous. However, contrary to our findings, the earlier parity-to-insulin sensitivity relationship disappeared after adjustment for age, pregestational BMI, and weight gain, which are relevant confounders that are strongly related to worsening insulin sensitivity. These data could lead us to speculate on whether there is a true effect on this relationship or whether this could arise solely from confounding. Importantly, adjustment of our analyses for these factors simultaneously with other behavioral factors (including educational level, smoking, PA, and eating habits) did not change the associations observed. Therefore, we can state that our results are robust under different modeling approaches. The apparent disagreement with the aforementioned findings could be due to differences in study design, population characteristics, and the methodologies used for evaluating IR, as well as the fact that in our study, the assessment period was in the first trimester.

Interestingly, supporting our results at least in part, the study by Seghieri et al (19) found that in comparison with the initial pregnancy, insulin sensitivity significantly decreased in the subsequent pregnancy. Similarly, Skajaa et al (6), studying pregnant women with type 1 diabetes, reported that daily insulin requirements from week to week increased significantly throughout pregnancy with increasing parity after adjusting for BMI, age, and prepregnancy glycated hemoglobin A_1c. This last study also revealed that the individual total insulin requirement in the woman's first pregnancy increased with each pregnancy during pregnancy but did not in nonpregnant status. This reaffirms, in part, our results and further strengthens the hypothesis that parity per se negatively affects insulin sensitivity during pregnancy, thereby increasing insulin requirement.

In addition, our findings are also consistent with the evidence from published studies indicating a relationship between multiparity and GD (11, 13), and most researchers report increased risk of diabetes in the later life of women with high parity (8, 9). However, it has been suggested that a woman would need to have at least 4 pregnancies for the risk of diabetes to be affected (8, 9). Unfortunately, our data do not permit us to test the effect of having more than 3 (n = 3) or 4 (n = 1) children on first-trimester IR because of the small number of women with these pregnancies; however, it is worth examining in future research.

It is well known that being OWO before and during pregnancy predisposes the woman to a higher metabolic dysregulation in pregnancy (38) including early IR, as observed in our study. According to our results, OWO women were more IR at the start of their pregnancy compared with NW women regardless of parity. This finding is in accordance with previous reports (22, 38). An original aspect of this study that deserves attention is the association of the combination of parity and OWO with IR in early pregnancy, which has not been previously researched. According to our study findings, nulliparous pregnant women with an ep-BMI of 25.0 or greater had an almost 3 times higher risk of IR compared to NW nulliparous women. It is interesting to note that OWO multiparous pregnant women had a 6 times higher risk of IR. Moreover, in this vulnerable group, the risk of IR showed an increasing trend with increasing parity. The above data suggest that the combination of parity and maternal OWO may result in additive adverse effects that contribute to a worse IR early in pregnancy.

The relationship between parity and maternal IR early in pregnancy is complex and still not fully understood. This probably reflects a combination of sociodemographic/environmental factors and placental hormone changes due to repeated pregnancies that result indirectly or directly in a progressive worsening of IR. We found, for example, that multiparous pregnant women were older and were less likely to have a university education and more likely to belong to a higher social class than nulliparae. In the present study, the last 2 factors were significantly related to higher ep-BMI, which in turn were independently and positively associated with early IR.

Multiparity is also an independent predictor for obesity (39). About three-quarters of women are unable to reach their prepregnancy weight 1 year after delivery (40), this leads to interpregnancy accumulation of adipose tissue, thereby contributing to IR in successive pregnancies and beyond (41). Nevertheless, our data do not allow us to draw conclusions about this issue. It has also been argued that, just like outside pregnancy, obesity in pregnancy promotes a high production of proinflammatory factors and oxidative stress due to the accumulation of adipose tissue macrophages, which induces IR (39, 42). Thus, obese multiparous pregnant women could have even greater metabolic consequences, including IR, as confirmed in our study. This is probably explained by either the greater adiposity, repeated episodes of inflammatory response, or a combination of these. Furthermore, regardless of obesity, an increase in the production of placental hormones, such as progesterone, estrogen, and placental lactogens, in pregnancy to accommodate the growing fetus (43) leads to a relative IR. This effect can accumulate, especially for multiparous women (39).

Taken together, our study offers new knowledge about the role of multiparity in maternal IR in subsequent pregnancies, which provides an opportunity for an early targeted intervention. However, it is necessary to consider several limitations when interpreting our findings. First, our population consisted of healthy pregnant women living in the Mediterranean region with specific sociodemographic and healthy Mediterranean lifestyle traits, which prevents the generalization of the findings to other populations. In addition, due to limited resources, it was not feasible to use the “gold-standard” hyperinsulinemic-euglycemic clamp test in early pregnancy to quantify IR (33). However, the HOMA-IR is a noninvasive, cost-effective, and simple surrogate measure of this parameter widely used in large epidemiologic studies and has also been validated in studies with pregnant women (44). Moreover, the fact that we did not have information on stillbirths due to early miscarriages and intrauterine fetal demise could have induced a classification bias. Last, previous findings support the importance of other potential factors mediating the recurrence of GD, such as the interpregnancy interval and percentage of body fat or fat distribution pattern through successive pregnancies (45); not having these data is, therefore, another potential weakness.

In summary, our study suggests that multiparity may have a negative effect on maternal IR, which is already detectable during the early stages of pregnancy, and that OWO appears to further aggravate this relationship. Therefore, it is strongly recommended that all pregnant women who have given birth previously, especially those with overweight or obesity, undergo early IR screening during subsequent pregnancies. Undoubtedly, identifying elevated IR at a very early stage in these high-risk pregnant women would make it possible to implement earlier interventions involving lifestyle modifications and/or pharmacological treatment to avoid future complications that could affect both the mother and baby.

Acknowledgments

We would like to thank all the volunteers for their participation and all staff for their contribution to the ECLIPSES trial. We would also like to thank the Jordi Gol Research Institute in Primary Care (Instituto de Investigación en Atención Primaria; IDIAP).

Funding

The ECLIPSES trial was supported by the financially by the Spanish government's official funding agency for biomedical research, Instituto de Salud Carlos III (ISCIII), through the Fondo de Investigación para la Salud (FIS) (Nos. PI12/02777 and PI17/01754) and co-funded by the European Union ERDF/ESF, “A way to make Europe”/“Investing in your future”. These funding bodies played no part in designing the study, collecting and interpreting the data, or deciding to publish. A.D. is a Serra Hunter Fellow, Spain.

Author Contributions

V.A. designed and conducted the research and performed data curation. A.D. and E.M. analyzed the data and wrote the article. All authors revised the manuscript for important intellectual content and read and approved the final version. The corresponding author attests that all listed authors meet authorship criteria and that no one who meets the criteria has been omitted. V.A. is the guarantor of this work and as such has had full access to all study data and takes responsibility for their integrity and for the accuracy of the data analysis.

Disclosures

The authors declare that they have no conflict of interest.

Data Availability

The data sets generated and/or analyzed during the present study are not publicly available because of subject confidentiality; however, they are available from the corresponding author on reasonable request.

Abbreviations

 
• ECLIPSES

  Ensayo CLInico Para Suplementar con hierro a EmbarazadaS

 
• ep-BMI

  early-pregnancy body mass index

 
• GD

  gestational diabetes mellitus

 
• HOMA-IR

  homeostasis model assessment of insulin resistance

 
• MET

  metabolic equivalent

 
• NW

  normal weight

 
• OWO

  overweight/obesity

 
• PA

  physical activity

 
• rMedDiet

  Relative Mediterranean Diet

 
• RR

  relative risk

 
• SES

  socioeconomic status

 
• T

  tertile

 
• VIF

  variance inflation factor