Plasma glycemic measures and fecundability in a Singapore preconception cohort study

Abstract

Objective

To examine the association between plasma glycemia in women attempting to conceive and fecundability, as measured by time-to-pregnancy (TTP).

Design

Prospective preconception population-based study.

Setting

KK Women’s and Children’s Hospital, Singapore.

Patient(s)

Asian preconception women, aged 18-45 years, attempting conception for ≤12 cycles at study entry.

Intervention(s)

None.

Main Outcome Measure(s)

We ascertained TTP within a year of glycemic assessment, in menstrual cycles. We estimated fecundability ratios (FRs) and 95% confidence intervals (CIs) using discrete-time proportional hazards models, adjusting for age, ethnicity, education, body mass index and cycle regularity, and accounting for left-truncation and right censoring.

Result(s)

We studied a population sample of 766 women from the Singapore PREconception Study of long-Term maternal and child Outcomes prospective cohort. Compared to women with normoglycemia, women with dysglycemia (prediabetes and diabetes, defined by American Diabetes Association) had a lower FR (0.56, (95% confidence interval 0.38, 0.83)). Compared with the respective lowest quintiles, women in the highest quintile of fasting glucose (≥5.1 mmol/L) had a FR of 0.60 (0.42,0.86), while women in the highest 2-hour post-load glucose quintile (≥6.9 mmol/L) had a FR of 0.66 (0.45,0.97). Overall, the FRs decreased generally across the range of fasting and 2-hour plasma glucose. Glycated hemoglobin was not associated with fecundability.

Conclusion(s)

Increasing preconception plasma glucose is associated with reduced fecundability, even within the normal range of glucose concentrations.

Introduction

Worldwide, approximately 50% of countries recorded total fertility rates (TFRs) below the replacement rate of the nation (1). This is pertinent in Singapore, where TFR fell from 1.96 in 1988 to 1.14 in 2018, which is now far below the replacement rate of 2.05 (2). Coupled with an increasing life expectancy, the country is experiencing rapid population aging that has profound economic and social implications (1). Both type 1 diabetes (T1D) or type 2 diabetes (T2D) have been linked with reproductive abnormalities in women, such as delayed menarche, menstrual irregularities, infertility and premature menopause (3). The sharp rise of T2D incidence among young adults globally may lead to the possibility of more reproductive aged women encountering diabetes-related fertility issues (4). The potentially modifiable nature of glycemia or T2D risk underscores the need for increased scientific research and public health focus on this area. There is an urgent call for efforts to break the intergenerational cycle of diabetes and fertility problems (5).

High risk for future T2D relates to three distinct prediabetes states – impaired fasting glucose (IFG) and impaired glucose tolerance (IGT) based on oral glucose tolerance test (OGTT), and elevated glycated hemoglobin (HbA1c) (6, 7). It has been argued that classifying an individual as prediabetes or diabetes would have neglected the differential pathophysiological roles of IFG and IGT, as well as the possible continuum of risk across the normal glycemic range (6, 8). Thus, it has been proposed that analyzing glycemic measures, including fasting plasma glucose (FG), 2-h post-load plasma glucose (2hPG) and HbA1c in a continuous form may perform better, and have a greater statistical power, to predict the risk rather than as categorical variables based on specific thresholds (8).

Recently, an increasing number of studies have focused on examining the links of prediabetes and T2D with female fecundability (the per-cycle probability of conception) (5, 9–11). However, evidence of their association was limited and findings were inconsistent. In particular, it remains unclear whether prediabetes can influence fecundability in women. Also, there is a lack of comprehensive assessment of women’s glycemic status preconception – a window that could influence downstream reproductive and developmental outcomes for both mother-offspring and intergenerational health (12). Using data from the Singapore PREconception Study of long-Term maternal and child Outcomes (S-PRESTO), we explored the relationships between women’s preconception plasma glycemic measures (FG, 2hPG and HbA1c) and fecundability, measured by time-to-pregnancy (TTP) in menstrual cycles. We hypothesized that increasing plasma glycemia measured as FG, 2hPG and HbA1c as well as having a dysglycemic status (prediabetes and diabetes) would be associated with a reduction in fecundability.

Materials and Methods

S-PRESTO (ClinicalTrials.gov, NCT03531658) is a prospective preconception cohort study that was designed to examine the long-term environmental effects of preconception exposures occurring before and during early pregnancy on mother-offspring metabolic and mental health. Asian women of Chinese, Malay or Indian ethnicity attempting to conceive within the next 12 months and aged between 18–45 years were recruited from the general population of Singapore. We excluded women with known T1D or T2D, and those who had been taken anticonvulsant medication, oral steroids or received assisted fertility treatment in the past one month. The study was conducted according to the guidelines laid down in the Declaration of Helsinki. The Singhealth Centralized Institute Review Board approved the study protocol (reference 2014/692/D). All participants provided written informed consent.

Study procedure

Details of the study protocol has been described elsewhere (13). Briefly, at baseline, women completed questionnaires on socio-demography, health history, menstrual characteristics and lifestyle factors via face-to-face interview. Research staff measured height, weight, waist and hip circumferences of women in the S-PRESTO cohort center at KK Women’s and Children’s hospital (KKH). Women were followed at a subsequent month where the menstrual characteristics were recorded. We provided women with home urinary pregnancy test kits (Biotron Diagnostics, USA) detecting the beta subunit of human chorionic gonadotropin and instructions on their use. Women were instructed to perform a pregnancy test if their menstrual periods were late for 3-4 days, or 2 weeks after unprotected intercourse. When a positive result on the pregnancy test kit was found, women were scheduled for an ultrasound scan to confirm pregnancy viability. In the absence of any update within 6, 9 and 12 months from the recruitment visit, research staff conducted a follow-up survey by telephone to track women’s pregnancy status. All women were followed for up to one year while attempting to conceive.

Plasma glucose and HbA1c

At the recruitment (baseline) visit, women underwent an OGTT with a 75g glucose load after an overnight fast. Plasma FG, 2hPG and HbA1c at baseline were measured using the ARCHITECT c8000 Clinical Chemistry Analyzer (Abbott Laboratories), which was accredited by the National Glycohemoglobin Standardization Program. Dysglycemia was defined as newly identified prediabetes and diabetes, according to the updated American Diabetes Association criteria (7). Prediabetes was defined as having an FG 5.6 to 6.9 mmol/L (IFG) or 2hPG 7.8-11.0 mmol/L (IGT) or HbA1c 5.7-6.4%. T2D was defined as FG ≥7.0 mmol/L or 2hPG ≥11.1 mmol/L or HbA1c ≥6.5%. Owing to the restricted number of women with T2D (n=16), these women were grouped together with those having prediabetes in this study. The remainder were classified as normoglycemia.

Outcome

TTP was defined as the number of menstrual cycles required to achieve a pregnancy over one year of follow-up. Pregnancy was determined by a positive urinary pregnancy test confirmed by the presence of an intrauterine gestational sac based on an ultrasound scan after 6 weeks of gestation. We calculated the interval between the dates of LMP at recruitment (baseline) and before conception (in those pregnant within one year) or last follow-up call (if not pregnant or withdrew from study). The interval was then converted to cycles by dividing with the average cycle length, which was derived from the reported minimum and maximum lengths of their usual cycles at baseline. For women reported uncertainty of their cycle lengths or having irregular cycles, we used data from a follow-up questionnaire (self-reported cycle length or difference in LMP dates) to verify or estimate cycle lengths. Irregular cycle was defined as having menstrual periods varied by more than five days in the past six months. We asked women about the number of months of attempting to conceive at study entry. TTP was calculated as the total discrete cycles at risk of pregnancy: (months of conception attempt at study entry/ average cycle length) + [(date of LMP before conception or the most recent follow-up) - (date of LMP at recruitment)]/ average cycle length. For women who became pregnant, one more conception cycle was added (14).

Statistical analysis

We compared differences in women’s characteristics by glycemic status using Pearson’s chi-squared test for categorical variables and Mann-Whitney test for continuous variables. We used discrete-time proportional hazards model, which analyzed TTP as a discrete scale (number of menstrual cycles), to estimate hazards ratio of fecundability (termed as fecundability ratio (FR)) and 95% confidence interval (CI) (15, 16). The FR represents the cycle-specific probability of conception in one group of women, relative to a control group. Unlike hazards ratio for mortality or morbidity, in which a ratio <1.0 shows a benefit, a FR <1.0 is undesirable. Thus, a FR <1.0 indicates reduced fecundability (longer TTP), while a FR >1.0 indicates increased fecundability (shorter TTP). To account for left truncation, we based risk sets only on observed cycles at risk, i.e. cycles of pregnancy attempt while participating in the study. For example, if a woman had been trying to conceive for four cycles at study entry and then reported a pregnancy after seven cycles of total attempt time, only three cycles (i.e. 5^th to 7^th cycles) as observed in the study contributed to the analysis. Women were censored if they (i) had not conceived after 12 months from the recruitment, (ii) initiated fertility treatment, (iii) reported no longer trying to conceive (iv) or was lost to follow up, whichever occurred first.

In view of the possibility that women who had been attempting to conceive for a long period at study entry might have adjusted their behavior and potentially influenced plasma glycemia, we performed main analysis by restricting samples to women with pregnancy attempt of ≤12 months at study entry. This would help to eliminate potential reverse causality and also, to exclude potential cases with underlying pathologies in female and male fertility. We constructed separate models for each glycemic measure to assess fecundability, analyzed crude and with confounders adjustment. A pool of potential confounders was selected a priori from the literature (5, 9–11), based on clinical understanding and by using a directed acyclic graph. Potential confounders which were adjusted for in the model included age (<30, 30-34, ≥35 years), ethnicity (Chinese, Malay, Indian, mix), education (below tertiary, tertiary and above), BMI (<18.5, 18.5-24.9, 25-29.9, ≥30 kg/m²) and cycle regularity (regular, irregular). Potential confounders which altered the effect estimates minimally were not included. Glycemic variables were analyzed in the forms of dyslycemia, quintiles of FG, 2hPG and HbA1c. As the precision level of glycemic measure was limited to decimal 0.1 mmol/L, this resulted an uneven distribution of women across quintiles.

We tested the possibility of a non-linear relationship between plasma glycemia and fecundability by adding glucose squared to the model (plasma glucose was treated in continuous form), and by using the restricted cubic spline with five knots located at 5^th, 25^th, 50^th, 75^th and 95^th percentiles. The reference point was determined based on the median level of plasma glucose among normoglycemic women. We performed additional analysis to examine the association between plasma glycemia and fecundability by including women with pregnancy attempt of >12 months at study entry. We also performed sensitivity analyses to estimate fecundability and TTP by treating the event of interest as a live birth pregnancy outcome (n=327). In order to avoid findings associated with an elevated plasma glycemia being driven mainly by extremes of glycemia (T2D) but not prediabetes, we excluded women with T2D (n=16) from the analysis. We also excluded women with self-reported known polycystic ovarian syndrome (PCOS; n=10) from the analysis. Statistical analyses were conducted using SPSS Statistics Version 19.0 (IBM Corp, Armonk, NY, USA) or Stata Statistical Software, Release 13 (StataCorp, College Station, TX, USA).

Results

Among 1032 women recruited into the S-PRESTO study, 136 were excluded due to invalid and/ or incomplete data. Of the remaining 896 women, 130 reported pregnancy attempts of >12 months before participating in this study. We included a final sample of 766 women with pregnancy attempt of ≤12 months at study entry in the main analysis (Figure 1). Excluded women (n=266) were similar in ethnicity, parity and cycle length, but more likely to be older, attain lower education, be obese and have irregular menstrual cycles, compared with those included women (Supplemental Table 1, available online).

Figure 1.

Flowchart of the study

The 766 women in the study contributed to 8875 cycles and 362 pregnancies within 12 months of recruitment. There were 185 (24.2%) and 310 (40.5%) women who conceived within 6 and 12 cycles of follow-up, respectively. Period of attempting to conceive at study entry was at the median of 0.97 cycles (interquartile range 0 to 4.7). We censored 404 women (52.7%) who did not conceive after 12 months of follow-up (n = 374), who started fertility treatment (n = 13), who self-withdrew or were lost to follow-up (n = 17) prior to 12 months.

Baseline characteristics of women classified as either normoglycemic or dysglycemic are shown in Table 1. Overall, 103 women (13.4%) were found to have dysglycemia. These women were more likely to be of Malay ethnicity (23.3% vs. 12.7%), attained lower educational levels (53.4% vs. 29.9% with below tertiary level), to be overweight (31.1% vs. 13.1%) and obese (29.1% vs. 8.0%), to have waist-hip ratio ≥0.85 (66.7% vs. 50.9%) and irregular menstrual cycles (44.7% vs. 30.3%), compared to those with normoglycemia. No differences in age, parity, marital status, cycle length, cigarette smoking, alcohol intake, folic acid supplement intake and physical activity were observed between both groups of women.

TABLE 1. Characteristics between women of normoglycemic and dysglycemic status from the S-PRESTO study, 2015-2018 (n=766).

Characteristics	Total (n=766)	Normoglycemic (n=663)	Dysglycemica (n=103)	pb
Age, n (%)				0.509
<30 years	298 (38.9)	258 (38.9)	40 (38.8)
30-34 years	375 (49.0)	328 (49.5)	47 (45.6)
≥35 years	93 (12.1)	77 (11.6)	16 (15.5)
Ethnicity, n (%)				0.034
Chinese	566 (73.9)	498 (75.1)	68 (66.0)
Malay	108 (14.1)	84 (12.7)	24 (23.3)
Indian	68 (8.9)	59 (8.9)	9 (8.7)
Mix	24 (3.1)	22 (3.3)	2 (1.9)
Parity				0.274
0	488 (63.8)	428 (64.7)	60 (58.3)
1	215 (28.1)	184 (27.8)	31 (30.1)
≥2	62 (8.1)	50 (7.6)	12 (11.7)
Highest education, n (%)				<0.001
Below tertiary	253 (33.0)	198 (29.9)	55 (53.4)
Tertiary and above	513 (67.0)	465 (70.1)	48 (46.6)
Marital status, n (%)				0.084
Married	747 (97.5)	645 (97.3)	102 (99.0)
Unmarried	17 (2.2)	17 (2.6)	0
Divorced/ separated	2 (0.3)	1 (0.2)	1 (1.0)
Body mass index, n (%)				<0.001
Underweight <18.5 kg/m2	68 (8.9)	62 (9.4)	6 (5.8)
Normal 18.5-24.9 kg/m2	496 (64.8)	461 (69.5)	35 (34.0)
Overweight 25-29.9 kg/m2	119 (15.5)	87 (13.1)	32 (31.1)
Obese ≥30.0 kg/m2	83 (10.8)	53 (8.0)	30 (29.1)
Waist-hip ratio, n (%)				0.008
<0.80	137 (17.9)	127 (19.2)	10 (9.8)
0.80-0.84	222 (29.1)	198 (29.9)	24 (23.5)
≥0.85	405 (53.0)	337 (50.9)	68 (66.7)
Cycle regularity, n (%)				0.005
Regular	519 (67.8)	462 (69.7)	57 (55.3)
Irregular	247 (32.2)	201 (30.3)	46 (44.7)
Cycle length, days	29.5 (29.0-32.5)	29.5 (29.0-32.0)	30.0 (29.0-37.5)	0.232
Cigarette smoking, n (%)				0.890
No	742 (96.9)	642 (96.8)	100 (97.1)
Yes	24 (3.1)	21 (3.2)	3 (2.9)
Alcohol intake, n (%)				0.142
No	235 (30.7)	197 (29.7)	38 (36.9)
Yes	531 (69.3)	466 (70.3)	65 (63.1)
Folic acid supplement intake, n (%)				0.244
No	383 (50.0)	326 (49.2)	57 (55.3)
Yes	383 (50.0)	337 (50.8)	46 (44.7)
Physical activityc, n (%)				0.245
Inactive	110 (14.4)	92 (13.9)	18 (17.5)
Minimally active	394 (51.5)	337 (50.9)	57 (55.3)
HEPA active	261 (34.1)	233 (35.2)	28 (27.2)
Fasting plasma glucose, mmol/L	4.7 (4.5-5.0)	4.7 (4.5-4.9)	5.2 (4.8-5.6)	<0.001
2-h post-load plasma glucosed, mmol/L	5.7 (5.0-6.6)	5.5 (4.9-6.1)	8.5 (7.8-9.6)	<0.001
Plasma HbA1c, %	5.1 (4.9-5.2)	5.1 (4.9-5.2)	5.4 (5.1-5.7)	<0.001
Attempt time at study entry, cycles	1.0 (0-4.7)	1.0 (0-4.3)	2.1 (0-5.6)	0.034

Note: Values are presented in n (%) for categorical variables and medians (25^th – 75^th percentiles) for continuous variables. Total sample size does not always equal to 766 due to the missing values. S-PRESTO = Singapore PREconception Study of long-Term maternal and child Outcomes; HEPA = health enhancing physical activity; HbA1c = glycated haemoglobin.

^aDysglycemia is defined as women with prediabetes and type 2 diabetes, according to 2020 American Diabetes Association criteria.

^bBased on Pearson’s chi-squared test for categorical variables and Mann-Whitney test for continuous variables.

^cClassified based on the International Physical Activity Questionnaire guidelines for data processing and analysis (17).

^dFrom 25g oral glucose tolerance test.

Associations between plasma glycemic measures and fecundability of women are presented in Table 2. In comparison to women with normoglycemia, women with dysglycemia had a lower FR of 0.56 (95% CI 0.38, 0.83). For FG, women in the highest quintile (≥ 5.1 mmol/L) showed the lowest FR of 0.60 (0.42, 0.86). Similarly, for 2hPG, women in the highest quintile (≥ 6.9 mmol/L) showed the lowest FR of 0.66 (0.45, 0.97). Overall, the FRs decreased generally across the range of FG (Figure 2A) and 2hPG, although there was a slight inflexion in the line graph observed at the lower range of 2hPG (Figure 2B). Using the restricted cubic spline and the glucose squared term which appeared to be non-significant, our results demonstrated the absence of non-linearity between plasma glucose and fecundability. No association was found between HbA1c and fecundability.

TABLE 2. Associations between plasma glycemic measures and fecundability in women from the S-PRESTO study, 2015-2018 (n=766).

				Unadjusted		Adjusteda
Glycemic measures	n	Pregnancies	Cycles	FR	95% CI	FR	95% CI
Glycemic status
Normoglycemia	663	331	2381	1.00	(ref.)	1.00	(ref.)
Dysglycemiab (prediabetes & diabetes)	103	31	213	0.59	0.41, 0.85	0.56	0.38, 0.83
Fasting plasma glucose, mmol/L
Q1 (≤4.4 mmol/L)	161	90	724	1.00	(ref.)	1.00	(ref.)
Q2 (4.5-4.6 mmol/L)	155	74	446	0.87	0.64, 1.18	0.82	0.60, 1.12
Q3 (4.7-4.8 mmol/L)	194	94	628	0.85	0.64, 1.14	0.83	0.62, 1.11
Q4 (4.9-5.0 mmol/L)	110	50	415	0.78	0.55, 1.10	0.77	0.54, 1.09
Q5 (≥5.1 mmol/L)	146	54	381	0.65	0.47, 0.92	0.60	0.42, 0.86
2-h post-load plasma glucose, mmol/L
Q1 (≤4.7 mmol/L)	149	79	612	1.00	(ref.)	1.00	(ref.)
Q2 (4.8-5.4 mmol/L)	158	86	619	1.09	0.81, 1.49	1.11	0.81, 1.51
Q3 (5.5-5.9 mmol/L)	152	71	534	0.85	0.61, 1.17	0.88	0.62, 1.23
Q4 (6.0-6.8 mmol/L)	164	76	453	0.87	0.64, 1.20	0.87	0.63, 1.21
Q5 (≥6.9 mmol/L)	143	50	376	0.63	0.44, 0.90	0.66	0.45, 0.97
Plasma HbA1c, %
Q1 (≤4.9%)	206	108	750	1.00	(ref.)	1.00	(ref.)
Q2 (5.0-5.0%)	133	59	451	0.81	0.60, 1.11	0.90	0.63, 1.28
Q3 (5.1-5.1%)	135	65	465	0.87	0.64, 1.19	0.83	0.59, 1.16
Q4 (5.2-5.3%)	179	88	578	0.95	0.72, 1.26	1.01	0.74, 1.36
Q5 (≥5.4%)	113	42	350	0.69	0.48, 0.99	0.73	0.48, 1.10

Note: Data were analyzed using discrete-time proportional hazards models. S-PRESTO = Singapore PREconception Study of long-Term maternal and child Outcomes; FR = fecundability ratio; CI = confidence interval; HbA1c = glycated hemoglobin.

^aAdjusted for age, ethnicity, education, body mass index, cycle regularity.

^bDysglycemia was defined based on 2020 American Diabetes Association criteria (fasting plasma glucose ≥5.6 mmol/L or 2-h post-load plasma glucose ≥7.8 mmol/L or glycated hemoglobin ≥5.7%).

Figure 2.

Associations of (2A) fasting plasma glucose and (2B) 2-hour post-load plasma glucose with fecundability ratio. Plasma glucose levels are divided into quintiles, with the first quintile serves as the reference. The line graphs are adjusted for age, ethnicity, education, body mass index and cycle regularity. Solid line represents fecundability ratio and error bars denote 95% confidence intervals.

When women with pregnancy attempt of >12 months at study entry were included in the analysis (Supplemental Table 2, available online), those with dysglycemia remained to have a reduced FR of 0.61 (0.38, 0.99). Women in the highest quintile of FG also remained to have a reduction in FR of 0.57 (0.40, 0.81). However, 2hPG was not significantly associated with fecundability. In the sensitivity analysis, when the event of interest was a pregnancy resulting in live birth (Supplemental Table 3, available online), no substantial changes in the FRs were observed. Findings remained similar when women with diabetes (Supplemental Table 4, available online) and PCOS (Supplemental Table 5, available online) were excluded. Reduced fecundability was consistently shown in women with dysglycemia and those in the highest quintiles of FG, but not for 2hPG.

Discussion

In this prospective cohort study involving reproductive aged Asian women who were planning conception, we examined the associations of plasma glycemic measures, including FG, 2hPG and HbA1c, with fecundability, as measured by TTP in cycles. These women were apparently healthy at recruitment with no awareness of their glycemic status. We found that both newly diagnosed prediabetes and T2D were associated with a reduction in fecundability. In general, worsening dysglycemia as indicated by increasing plasma glucose, even within the normal range, was associated with reduced fecundability (delayed TTP), and this was particularly evident for increasing FG. HbA1c, however, was not associated with fecundability. These findings suggest that monitoring and optimizing plasma glucose during preconception have the potential to shorten TTP and improve fecundability in women of reproductive age.

Our finding of delayed TTP in women with high FG is supported by a recent nationwide study in China that involved 2.2 million couples, reporting that IFG and diabetes (defined based on a single FG test) were associated with prolonged TTP in women attempting their first pregnancy (14). The authors demonstrated that the optimal FG for greatest fecundability was 3.9-4.9 mmol/L, thereafter, fecundability started to decline (14). The results are reasonably consistent with our present analysis, showing that an FG of 5.1 mmol/L and above was associated with reduced fecundability. Further, a general decline in fecundability was observed across the range of FG. Together, these findings indicate that women may experience a reduced fecundability with increasing FG, even at the level of normal range below specific clinical threshold.

While Zhao et al. (14) only focused on FG, we additionally examined 2hPG and HbA1c, which allowed us to interrogate which glycemic biomarkers could better predict fecundability. In this study, we provided new evidence suggesting that 2hPG has a weaker influence on fecundability than FG, while HbA1c was not useful in predicting fecundability, despite the expected correlations between FG, 2hPG and HbA1c (FG & 2hPG: r=0.71; FG & HbA1c: r=0.79; 2hPG & HbA1c: r=0.69; all p<0.001). This is probably because 2hPG could be subject to greater within-person variability (18); while HbA1c had poorer sensitivity for detecting dysglycemia (19, 20). We also found that the association between 2hPG and fecundability was attenuated when women with pregnancy attempt of >12 months at study entry were included in the analysis. However, FG was consistently associated with fecundability. This suggests that testing FG alone is acceptable without the need to undergo an OGTT in the management of subfertility, although replication in other large studies involving different populations are required. In contrast, a multicenter study in United States (n=501) reported that diabetes in preconception women was not associated with fecundability, although this could be explained by the very low number of women with diabetes (n=6) or treatment received (10). Nonetheless, an OGTT may be recommended in high risk women prior to conception for reasons other than the management of subfertility, such as the detection of undiagnosed T2D which is known to be associated with poor pregnancy and offspring outcomes if poorly controlled prior to conception.

There are potential biological mechanisms underlying the postulated association between diabetes and fecundability. Insulin resistance (IR), a driving factor that leads to prediabetes and T2D, has been shown to play a key role in the pathogenesis of female fertility impairment. As demonstrated in female mice, IR contributes to oxidative stress and disrupts mitochondrial function in oocytes, consequently impairing oocyte size and maturation (21). Similarly, IR and hyperinsulinemia associated with PCOS affect the developmental potential of immature oocytes, as indicated by impaired oocyte maturation and fewer fertilized oocytes (22), leading to decreased ovulation and conception rates per cycle (23). Even in the absence of PCOS, elevated glucose level is potentially impairing female fertility (14). In addition, the hypothalamic– pituitary–ovarian axis that serves as the main regulator of menstrual cycle may be altered by glucose levels and thus, affecting fecundability (3). This supports our observation of a higher likelihood of cycle irregularity in women with dysglycemia compared to those without. Despite adjustment for cycle regularity, a higher plasma glucose remained associated with fecundability, as supported by Whitworth et al. (9). and Zhao et al. (14). This points to other postovulatory mechanisms, including fertilization and implantation, that may be associated with plasma glycemia. In particular, women with T2D were found to have reduced sexual desire and increased dyspareunia, thereby decreasing the frequency of sexual intercourse and reducing fecundability (24). Further studies should examine the association between plasma glycemia and these postovulatory mechanisms, in order to improve fecundability.

In our cohort, 41% of women spontaneously conceived after 12 cycles of pregnancy attempts, which is lower than the reported rates of more than 70% in other studies of natural conception (25–27). At the point of recruitment, 59% women had already been trying to conceive for variable lengths of time. There is likely a bias as most fertile women would not have had the time to consider joining this observational study before they conceived. In this study, the women recruited expressed their intention to conceive and were encouraged to engage in sexual intercourse 2-3 times per week; however, the frequency of sexual activity may have been low, which was potentially attributable to other issues such as a stressful lifestyle or health conditions. Therefore, we cannot exclude the possibility that some women might have temporarily stopped or delayed their conception attempts during the study without informing us, even during the follow-up phone interviews; thus, the number of cycles at-risk might be overestimated, resulting in a low conception rate. However, we did not collect data on sexual activity, fertility monitoring and pregnancy intention tracking, limiting our ability to verify these postulations. Apart from possible underlying subfertility issues among our women, the low conception rate could be attributable to the exclusion of biochemical pregnancies in defining pregnancy events, which was based on ultrasound evaluation with the presence of gestational sac. Nevertheless, the low conception rate in this study is consistent with the relatively low TFR in Singapore (2).

By comparing the difference between number of pregnancies and live births, the incidence of pregnancy loss in this study was lower (10%) than previous estimates in Singapore (14-25%) (28, 29). It might be expected that our study would underestimate the rate of pregnancy loss due to the missed cases of early pregnancy loss before 6 weeks gestation. Moreover, 88% of women recruited were below age 35 years, and thus likely to have a lower risk of miscarriage. Also, we observed that the LMP data that was required to generate TTP was more likely to be missing or invalid in women with miscarriages than those with live births, and thus excluded from the present analysis. Besides, there might have been cases where the women misreported their pregnancy status and were censored in the analysis due to the reluctance to inform the study team about a miscarriage.

The main strength of this study was the prospective cohort design that began preconception, where women attempting to conceive were followed up closely, with determination of pregnancy confirmed by an ultrasound scan, with minimal recall bias. Glycemic status was assessed by the OGTT gold standard, alongside an HbA1c measure, which enabled us to have a more comprehensive examination of women’s glycemic status. Both categorical and continuous measures of plasma glycemia were examined. To ensure the robustness of our results, we performed the analysis by accounting for left truncation to adjust the attempt time to conceive at study entry. In reality, preconception women engaging healthcare professionals comprise individuals with varied periods of pregnancy attempt. To understand whether bias occurred due to potential reverse causality or by having women who were simply not able to become pregnant spontaneously due to underlying pathological conditions in themselves or their partners, we performed additional analyses including women who reported having tried to become pregnant for >12 months at study entry. Dysglycemia and FG remained associated with fecundability, indicating infertility bias was unlikely to have played a role in this finding. We also conducted sensitivity analysis by targeting on the event of interest based on pregnancies resulting in a live birth and excluding women with self-reported PCOS. Similar findings were observed between plasma glycemia and fecundability. Unfortunately, we did not collect information related to phenotypic markers of PCOS, such as body hair and acne, that could have assisted in verifying our findings. Also, we were unable to exclude women with anovulatory cycles, despite adjusting for irregular menses. Time-varying variables such as cycle length was not collected consecutively through the period of pregnancy attempt, thus, the accuracy of self-reported cycle length and menstrual regularity at baseline was not able to be validated. Nevertheless, it has been reported that regardless of menstrual cycle disturbances (length or regularity), women with T1D and T2D have reduced fecundability (9).

We had no information on coital frequency and ovulation data (timing and the use of ovulation inducing drugs) to verify the cycles at-risk, which is the main limitation of this study. The assessment of cycles at-risk was based on the assumption that all enrolled women would always be intending to conceive, engage in intercourse and experience per-cycle ovulation over the entire study period. However, this may not be true for all women and potentially, introducing an error for TTP estimation. Thus, the present findings should be interpreted cautiously. The extent to which our findings could be generalizable to a wider population or other ethnicities remains to be established, as the current study was confined to planned pregnancies among Asian women in Singapore. An approximately 44% of pregnancies in Singapore are unplanned (30). However, a previous study indicated that pregnancy planning had no effect on the association between diabetes and fecundability (5). Though differences in characteristics such as age, education, BMI and cycle regularity were observed between excluded and included women in the main analysis, we had controlled all these variables in the analysis. Nonetheless, owing to the lack of data collected for male partners’ characteristics and semen quality, we were not able to adjust for these variables; this might potentially lead to overestimation of the association between plasma glycemia and fecundability in this study.

Conclusions

Higher plasma glycemia in the preconception period, especially FG, is associated with a reduction in fecundability and prolonged TTP. Early evaluation and optimization of FG to improve fecundability through healthy diet and lifestyle modification are acceptable interventions. Having more optimal glycemia during preconception may not only shorten TTP, but it may also shape the environment in which the embryo will develop with consequently improved offspring outcomes.