Associations of periconceptional oral contraceptive use with pregnancy complications and adverse birth outcomes

Abstract

Background

Periconceptional use of oral contraceptives (OCs) has been reported to increase risks of pregnancy complications and adverse birth outcomes, but risks are suggested to differ depending on the timing of discontinuation, amount of oestrogen and progestin content.

Methods

Prospective cohort study among 6470 pregnancies included in the PRegnancy and Infant DEvelopment (PRIDE) Study in 2012–19. Exposure was defined as any reported use of OCs within 12 months pre-pregnancy or after conception. Outcomes of interest were gestational diabetes, gestational hypertension, pre-eclampsia, pre-term birth, low birthweight and small for gestational age (SGA). Multivariable Poisson regression using stabilized inverse probability weighting estimated relative risks (RRs) with 95% CIs.

Results

Any periconceptional OC use was associated with increased risks of pre-eclampsia (RR 1.38, 95% CI 0.99–1.93), pre-term birth (RR 1.38, 95% CI 1.09–1.75) and low birthweight (RR 1.45, 95% CI 1.10–1.92), but not with gestational hypertension (RR 1.09, 95% CI 0.91–1.31), gestational diabetes (RR 1.02, 95% CI 0.77–1.36) and SGA (RR 0.96, 95% CI 0.75–1.21). Associations with pre-eclampsia were strongest for discontinuation 0–3 months pre-pregnancy, for OCs containing ≥30 µg oestrogen and for first- or second-generation OCs. Pre-term birth and low birthweight were more likely to occur when OCs were discontinued 0–3 months pre-pregnancy, when using OCs containing <30 µg oestrogen and when using third-generation OCs. Associations with SGA were observed for OCs containing <30 µg oestrogen and for third- or fourth-generation OCs.

Conclusions

Periconceptional OC use, particularly those containing oestrogen, was associated with increased risks of pre-eclampsia, pre-term birth, low birthweight and SGA.

Key Messages.

• Periconceptional use of oral contraceptives (OCs) has been associated with pregnancy complications and adverse birth outcomes, but risks are suggested to differ depending on timing of discontinuation, amount of oestrogen and progestin content.

• Any periconceptional use of OCs was associated with modestly increased risks of pre-eclampsia, pre-term birth and low birthweight.

• Associations with pre-eclampsia, pre-term birth and low birthweight were strongest when OCs were discontinued 0–3 months before pregnancy.

• High oestrogen doses were associated with pre-eclampsia, whereas low to moderate oestrogen doses were associated with adverse birth outcomes.

• First/second-generation progestins were associated with pre-eclampsia and third-generation progestins with all birth outcomes studied.

Introduction

With the increasing desire to protect against unplanned pregnancies, there is a global trend of increasing use of modern contraceptive methods.¹ The global prevalence of oral contraceptive (OC) use by women of reproductive age (15–49 years) was estimated to be 8.0% in 2019.² Note that high variation exists across regions, ranging from 1.9% in Middle Africa to 31.5% in Western Europe. Regardless, OCs are commonly used medication for which relatively rare adverse effects can have a large impact on public health. Disorders that have been associated with OC use include hypertension, arterial and thromboembolic events, dyslipidaemia and insulin resistance.^3–7 When used in the periconceptional period, OCs were reported to reduce the risk of gestational hypertension and to increase the risks of pre-eclampsia, low birthweight and pre-term birth.^7–11 However, these findings have not been reproduced consistently.^7,12–16

Despite a delay of 2–3 months in the return of fertility after discontinuation of hormonal contraceptive use, a systematic review and meta-analysis did not find a reduced fertility rate within the first 12 months of discontinuation.¹⁷ However, systematic and comprehensive data on pregnancy complications and birth outcomes are lacking. Variance in risks of adverse effects after periconceptional OC use has been suggested to be partially driven by the types of progestin and oestrogen components.^7,11,18 Therefore, we aimed to determine associations of common pregnancy complications and adverse birth outcomes with periconceptional OC use overall, and according to timing of discontinuation, amount of oestrogen and progestin content.

Methods

Data collection

Data were prospectively collected within the PRegnancy and Infant DEvelopment (PRIDE) Study—an ongoing prospective cohort among pregnant women in the Netherlands aimed at identifying factors that may affect maternal and child health during and after pregnancy.¹⁹ Pregnant women aged ≥18 years were recruited as early in pregnancy as possible by prenatal care providers, through leaflets distributed by the ‘Moeders voor Moeders’ (Mothers for Mothers) initiative and via advertisements in magazines and on social media platforms. Participants completed Web-based questionnaires at baseline (median gestational age = 10 weeks), in gestational weeks 17 and 34, and 2 and 6 months after the estimated date of delivery. In case of delivery before gestational week 34, adapted follow-up questionnaires were administered. Paper-based questionnaires were available for participants who could not or chose not to participate via the Internet (1.1% of the study population). Furthermore, consent was asked to extract data from obstetric records (consent rate 70.6%). The PRIDE Study was approved by the Regional Committee on Research Involving Human Subjects (CMO 2009/305) and all participants provided online or written informed consent.

Exposure assessment

Information on periconceptional OC (i.e. combined oestrogen–progesterone and progesterone-only pills) use was obtained from the baseline questionnaire and defined as any use in the year before or during pregnancy. This period was chosen because the effects of OCs may still be noticeable for ≤1 year.²⁰ Furthermore, we assumed that there would be no exposure to OCs beyond the first trimester of pregnancy. Participants were also asked the timing of OC discontinuation (4–12 months before pregnancy, 0–3 months before pregnancy or during pregnancy) and their OC brand/dosage from a list of all OCs available in the Netherlands. Each monophasic and biphasic OC was classified by the amount of oestrogen [no oestrogen (progestin-only pill), low to moderate dose (<30 µg) and high dose (≥30 µg)] and by progestin content [norethindrone (first-generation), norgesterel and/or levonorgesterel (second-generation), desogestrel and/or gestodene (third-generation) and drospirenone (fourth-generation)].²¹ Only a few participants used triphasic combination pills; these were excluded from the classification on the amount of oestrogen. Due to a relatively small number of first-generation OC users (44/7241, 0.6%), first- and second-generation progestin OC users were combined. The comparison group consisted of pregnancies without reports of OC use in the year before conception.

Outcomes of interest

Pregnancy complications of interest were gestational diabetes, gestational hypertension (>20 weeks of gestation) and late-onset pre-eclampsia (≥34 weeks; hereafter referred to as ‘pre-eclampsia’). Diagnostic criteria were based on the most recent Dutch guidelines²² and gestational hypertension and pre-eclampsia were mutually exclusive.²³ Early-onset pre-eclampsia was not considered because its pathophysiology differs from late-onset pre-eclampsia and the incidence was low in our study population.²⁴ Birth outcomes of interest were pre-term birth (<37 weeks), low birthweight (<2500 g) and small for gestational age (SGA). Newborns were considered to be SGA if the birthweight was below the 10th percentile for gestational age using the most recent Dutch reference curves.²⁵

Outcomes were primarily obtained from the prenatal questionnaires administered at gestational week 17 (gestational diabetes only) and week 34, and from the first post-partum questionnaire (all outcomes). These questionnaires were validated previously, with generally high levels of sensitivity (range 0.88–1.00) and specificity (range 0.99–1.00) for all outcomes, except for gestational hypertension (sensitivity 0.62, specificity 0.97).^26,27 In case of missing data on the outcomes due to skipped questionnaires or item non-response, outcome data were extracted from obstetric records whenever possible.

Covariates

We identified a minimally sufficient set of confounders and risk factors for each outcome through literature review and directed acyclic graphs (Supplementary Figures S1–S6, available as Supplementary data at IJE online).^28,29 The different sets included maternal age; race/ethnicity, which was operationalized as nationality [Dutch (participant and both parents born in The Netherlands) vs non-Dutch]; educational level [high (university or higher vocational education graduate) vs low/intermediate]; parity (0 vs ≥1 previous deliveries); diagnosis of polycystic ovarian syndrome; pre-existing depression; pre-existing anxiety; pregnancy intention; history of pre-term birth; smoking during pregnancy; and alcohol consumption during pregnancy.

Inclusion and exclusion criteria

We considered all pregnancies included in the PRIDE Study with an estimated date of delivery between 1 January 2012 and 31 December 2019 (N = 9054). Pregnancies missing information on periconceptional OC exposure, exposed to other methods of hormonal contraception (i.e. hormonal intrauterine devices, skin implants, injections, patches or vaginal ring), as a result of fertility treatment containing hormones, resulting in multiple gestations, which were ectopic or molar pregnancies, resulting in miscarriage or which were electively terminated were excluded. In addition, pregnancies among women with pre-existing diabetes mellitus were excluded from the analyses on gestational diabetes; pregnancies among women with chronic hypertension and those who used antihypertensives at baseline were excluded from the analysis on gestational hypertension; and pregnancies resulting in a stillbirth or neonatal death were excluded from the analyses on adverse birth outcomes.

Statistical analyses

All statistical analyses were performed in R/Rstudio. Maternal and pregnancy characteristics were reported as mean and standard deviation or number and percentage, as appropriate. Poisson regression models with a log link were fitted to estimate the crude relative risk (RR) for each exposure–outcome association^30,31 and 95% CIs with robust standard errors to account for participants with more than one pregnancy in the PRIDE Study.³²

To account for measured confounders, propensity scores were estimated and used as inverse probability of treatment weights (IPTWs). Propensity score models were fitted using generalized linear mixed-effects binomial regression models with a logit link (‘lme4’ package) and robust variance estimator for any vs no periconceptional OC use,³³ in which the minimally sufficient set of confounders specific for each outcome of interest were used as covariables to derive the propensity scores. Under the assumption that data were missing at random, missing data on covariables were handled using multiple imputation (fully conditional specification) with five iterations to create a data set without missing covariables (‘mice’ package).³⁴ Second-degree factorial polynomials for continuous variables (‘mfp’ package) and interaction terms between two covariables were included if these contributed statistically significantly to the model (P < 0.05).³⁵ For the calculation of all propensity scores, age was transformed to the third power and an interaction term between age³ and parity of ≥1 was included. To compare no periconceptional OC use to multiple exposure groups (i.e. timing of OC discontinuation, amount of oestrogen and progestin content), multinomial logistic regression (‘nnet’ package) was applied.^36,37

We also accounted for loss to follow-up by estimating separate propensity scores to calculate inverse probability of censoring weights (IPCWs) using a similar approach. Subsequently, stabilized IPTWs and IPCWs were calculated by multiplying the reciprocal of the conditional probability (i.e. the propensity score for the pregnancy’s actual exposure or censoring) with the marginal probability of the pregnancy’s actual exposure or censoring.^38–41 For each analysis, the final weight given to each pregnancy was the stabilized IPTW multiplied by the stabilized IPCW. Modified Poisson regression models were weighted by IPTW*IPCW to obtain the weighted and adjusted RRs (hereafter referred to as ‘adjusted RR’).

Additional analyses were performed to determine associations for exposure characteristics, namely the timing of discontinuation of periconceptional OC use, oestrogen dose and progestin content. Pregnancies with missing exposure details were excluded from the relevant propensity score calculation and final analyses. To compensate, we assumed that these exposures were missing at random for which the weights of non-missing OC users were adjusted proportionally. Crude RRs were only calculated for exposure characteristics with more than one event; adjusted RRs were only calculated for exposure characteristics with five or more events.

In exploratory analyses, we stratified by parity (0 vs ≥1 previous deliveries). To explore whether pre-eclampsia mediated the associations between periconceptional OC use and pre-term birth, we conducted a post hoc mediation analysis.⁴² Robust standard errors based on 100 simulations were used to define the 95% CIs. This approach was repeated to assess whether the associations for low birthweight may be due to pre-term birth.

Results

Of the 9054 pregnancies included in the PRIDE Study, 7241 (80.0%) were considered eligible for this study (Figure 1). A total of 3047 (42.1%) of these pregnancies were exposed to OCs in the periconceptional period. After excluding those without data on any outcomes of interest (10.6% of eligible pregnancies), 6470 pregnancies were included in the analyses with slightly lower numbers for each specific outcome (Figure 1). Pregnancies exposed to OCs in the periconceptional period were among women who were slightly younger (29.8 vs 31.3 years), were less likely to have a high level of education (72.6% vs 76.4%) or unintended pregnancy (5.4% vs 9.8%), had less often given birth before (35.9% vs 56.2%) and were more likely to have smoked (5.7% vs 3.5%) or consumed alcohol during pregnancy (13.6% vs 12.2%) compared with pregnancies not exposed to OCs in the periconceptional period (Table 1). In most pregnancies OCs were discontinued 4–12 months before pregnancy (61.5%), whereas in 36.2% OCs were discontinued 0–3 months before pregnancy, and in 2.2% in the first trimester of pregnancy.

Figure 1.

Flow chart of participation

Table 1.

Maternal and pregnancy characteristics of participants who used and did not use oral contraceptives in the periconceptional period

Characteristic	Periconceptional oral contraceptive use, n = 2715a		No periconceptional oral contraceptive use, n = 3755a
	Mean or frequency	SD or %	Mean or frequency	SD or %
Maternal age (years)	29.8	3.6	31.3	3.8
Dutch nationality	2430	92.2	3304	90.5
Parity ≥1	976	35.9	2111	56.2
High educational level	1922	72.6	2806	76.4
Pre-pregnancy BMI (kg/m2)	24.0	4.2	23.5	4.1
Pre-pregnancy smoking	350	13.2	359	9.7
Chronic conditions
Polycystic ovarian syndrome	26	1.0	106	2.8
Depression	35	1.3	61	1.6
Anxiety	22	0.8	33	0.9
Diabetes mellitus	1	0.0	5	0.1
Hypertension	4	0.1	6	0.2
Unintended pregnancy	147	5.4	364	9.8
Previous pre-term birth	77	2.8	137	3.6
Antihypertensive use at baseline	3	0.1	5	0.1
Alcohol during pregnancy	360	13.6	450	12.2

Data from the PRIDE Study, 2012–2019. Continuous variables are given as means and SDs; factors are given as frequencies or fractions (if exclusion criteria apply) and percentages. BMI, body mass index.

^aMissing data: Dutch nationality, n = 182 (2.8%); Educational level, n = 150 (2.3%); Pre-pregnancy BMI, n = 77 (1.2%); Pre-pregnancy smoking, n = 132 (2.0%); Pregnancy intention, n = 45 (0.7%); Alcohol during pregnancy, n = 140 (2.2%).

Pregnancy complications

Compared with no periconceptional OC use, any periconceptional OC use seemed to be associated with a slightly increased risk of pre-eclampsia (adjusted RR 1.38, 95% CI 0.99–1.93) but not with gestational diabetes (1.02, 0.77–1.36) and gestational hypertension (1.09, 0.91–1.31) (Table 2). Compared with pregnancies not exposed to OCs in the periconceptional period, pre-eclampsia occurred more often in pregnancies in which OCs were discontinued 0–3 months before pregnancy (1.56, 1.03–2.37), oestrogen doses ≥30 µg (1.53, 1.09–2.15) and first- or second-generation OCs (1.48, 1.04–2.11).

Table 2.

Associations between periconceptional oral contraceptive use and pregnancy complications

Periconceptional OC usea	Gestational diabetes					Gestational hypertension					Pre-eclampsia
	Total	n	%	cRR (95% CI)	aRR (95% CI)b	Total	n	%	cRR (95% CI)	aRR (95% CI)c	Total	n	%	cRR (95% CI)	aRR (95% CI)c
No periconceptional OC use	3748	139	3.7	Reference	Reference	3741	240	6.4	Reference	Reference	3748	67	1.8	Reference	Reference
Periconceptional OC use	2713	84	3.1	0.83 (0.64–1.09)	1.02 (0.77–1.36)	2706	226	8.4	1.30 (1.09–1.55)	1.09 (0.91–1.31)	2711	80	3.0	1.65 (1.20–2.28)	1.38 (0.99–1.93)
Timing of discontinuation
4–12 months pre-pregnancy	1641	53	3.2	0.87 (0.64–1.19)	1.11 (0.79–1.56)	1639	126	7.7	1.20 (0.97–1.47)	1.02 (0.82–1.28)	1640	41	2.5	1.40 (0.95–2.05)	1.12 (0.75–1.66)
0–3 months pre-pregnancy	973	26	2.7	0.72 (0.48–1.09)	0.78 (0.47–1.27)	968	86	8.9	1.38 (1.09–1.75)	1.13 (0.88–1.44)	972	34	3.5	1.96 (1.30–2.94)	1.56 (1.03–2.37)
During pregnancy	50	4	8.0	2.16 (0.83–5.60)	N/A	50	11	22.0	3.43 (2.01–5.86)	0.31 (0.04–2.15)	50	1	2.0	N/A	N/A
Amount of oestrogen
None	138	6	4.3	1.17 (0.53–2.61)	0.63 (0.26–1.51)	137	9	6.6	1.02 (0.54–1.95)	2.47 (0.83–7.32)	137	1	0.7	N/A	N/A
Low to moderate (<30 µg)	336	8	2.4	0.64 (0.32–1.30)	0.92 (0.45–1.88)	336	22	6.5	1.02 (0.67–1.56)	0.81 (0.52–1.28)	336	7	2.1	1.17 (0.54–2.52)	1.10 (0.48–2.51)
High (≥30 µg)	1946	58	3.0	0.80 (0.59–1.09)	1.05 (0.76–1.46)	1941	174	9.0	1.38 (1.16–1.69)	1.15 (0.95–1.40)	1945	68	3.5	1.96 (1.40–2.73)	1.53 (1.09–2.15)
Type of progestin
First/second-generation	1779	58	2.3	0.88 (0.65–1.19)	1.13 (0.82–1.57)	1775	152	8.6	1.33 (1.10–1.62)	1.12 (0.91–1.37)	1779	59	3.3	1.86 (1.31–2.62)	1.48 (1.04–2.11)
Third-generation	465	13	2.8	0.75 (0.43–1.32)	0.69 (0.39–1.22)	464	34	7.3	1.14 (0.81–1.61)	1.11 (0.78–1.57)	464	9	1.9	1.09 (0.54–2.16)	1.10 (0.55–2.20)
Fourth-generation	213	4	1.9	0.51 (0.19–1.36)	N/A	212	24	11.3	1.76 (1.19–2.62)	1.29 (0.84–1.97)	212	8	3.8	2.11 (1.03–4.34)	1.26 (0.60–2.63)

Data from the PRIDE Study, 2012–2019. aRR, adjusted relative risk; cRR, crude relative risk; n, number of events; N/A, not applicable; OC, oral contraceptive.

^aUp to 1 year before or during pregnancy.

^bAdjusted for maternal age, nationality, parity, pre-pregnancy smoking, pregnancy intention and polycystic ovarian syndrome diagnosis.

^cAdjusted for maternal age, nationality, educational level, parity, pre-pregnancy smoking and pregnancy intention.

Birth outcomes

We observed associations between any periconceptional OC use and pre-term birth (adjusted RR 1.38, 95% CI 1.09–1.75) and low birthweight (1.45, 1.10–1.92) but not SGA (0.96, 0.75–1.21) (Table 3). Pregnancies in which OCs were discontinued 0–3 months before pregnancy had increased risks of pre-term birth (1.55, 1.11–2.17) and low birthweight (1.88, 1.28–2.77), whereas discontinuation of OCs 4–12 months before pregnancy seemed to be associated with pre-term birth only (1.31, 0.99–1.71).

Table 3.

Associations between periconceptional oral contraceptive use and pregnancy outcomes

Periconceptional OC usea	Pre-term birth					Low birthweight					Small for gestational age
	Total	n	%	cRR (95% CI)	aRR (95% CI)b	Total	n	%	cRR (95% CI)	aRR (95% CI)c	Total	n	%	cRR (95% CI)	aRR (95% CI)c
No periconceptional OC use	3729	134	3.6	Reference	Reference	3607	93	2.6	Reference	Reference	3597	175	4.9	Reference	Reference
Periconceptional OC use	2695	153	5.7	1.58 (1.26–1.98)	1.38 (1.09–1.75)	2621	112	4.3	1.66 (1.26–2.17)	1.45 (1.10–1.92)	2614	121	4.6	0.95 (0.76–1.19)	0.96 (0.75–1.21)
Timing of discontinuation
4–12 months pre-pregnancy	1627	91	5.6	1.56 (1.20–2.02)	1.31 (0.99–1.71)	1584	60	3.8	1.47 (1.07–2.02)	1.24 (0.89–1.73)	1578	67	4.2	0.87 (0.66–1.15)	0.90 (0.67–1.22)
0–3 months pre-pregnancy	970	56	5.8	1.61 (1.19–2.18)	1.55 (1.11–2.17)	940	45	4.8	1.86 (1.31–2.63)	1.88 (1.28–2.77)	939	47	5.0	1.03 (0.75–1.41)	1.03 (0.72–1.46)
During pregnancy	50	1	2.0	N/A	N/A	49	1	2.0	N/A	N/A	49	3	6.1	1.26 (0.42–3.80)	N/A
Amount of oestrogen
None	136	5	3.7	1.02 (0.43–2.46)	0.35 (0.12–1.02)	128	4	3.1	1.21 (0.45–3.25)	N/A	127	10	7.9	1.62 (0.88–2.99)	1.56 (0.49–4.93)
Low to moderate (<30 µg)	333	24	7.2	2.01 (1.32–3.05)	1.75 (1.14–2.69)	324	22	6.8	2.63 (1.68–4.13)	2.43 (1.53–3.88)	324	27	8.3	1.71 (1.16–2.53)	1.81 (1.21–2.72)
High (≥30 µg)	1934	108	5.6	1.55 (1.21–1.99)	1.27 (0.98–1.64)	1883	75	4.0	1.54 (1.15–2.08)	1.25 (0.92–1.71)	1878	69	3.7	0.76 (0.57–0.99)	0.72 (0.53–0.96)
Type of progestin
First/second-generation	1669	99	5.6	1.56 (1.21–2.01)	1.25 (0.96–1.63)	1724	70	4.1	1.57 (1.16–2.14)	1.25 (0.91–1.72)	1720	70	4.1	0.84 (0.64–1.10)	0.80 (0.60–1.07)
Third-generation	430	30	6.5	1.81 (1.24–2.66)	1.83 (1.25–2.69)	445	25	5.6	2.18 (1.42–3.35)	2.29 (1.48–3.53)	443	33	7.4	1.53 (1.07–2.19)	1.45 (1.00–2.10)
Fourth-generation	203	9	4.2	1.18 (0.61–2.29)	0.83 (0.41–1.70)	202	6	3.0	1.15 (0.51–2.60)	0.75 (0.33–1.71)	201	5	2.5	0.51 (0.21–1.23)	0.34 (0.13–0.93)

Data from the PRIDE Study, 2012–2019. aRR, adjusted relative risk; cRR, crude relative risk; n, number of events; N/A, not applicable; OC, oral contraceptive.

^aUp to 1 year before or during pregnancy.

^bAdjusted for maternal age, nationality, educational level, parity, pre-pregnancy smoking, pregnancy intention and history of pre-term birth.

^cAdjusted for maternal age, nationality, educational level, parity, pre-pregnancy smoking, pregnancy intention, alcohol during pregnancy, pre-existing depression and pre-existing anxiety.

Low to moderate oestrogen doses (<30 µg) were associated with pre-term birth (adjusted RR 1.75, CI 1.14–2.69), low birthweight (2.43, 1.53–3.88) and SGA (1.81, 1.21–2.72) (Table 3). Use of OCs with oestrogen doses of ≥30 µg was linked to pre-term birth (1.27, 0.98–1.64) and to fewer cases of SGA (0.72, 0.53–0.96). Regarding progestin content, using third-generation OCs was associated with pre-term birth (1.83, 1.25–2.69), low birthweight (2.29, 1.48–3.53) and SGA (1.45, 1.00–2.10). Finally, fourth-generation OC users had a reduced risk of SGA (0.34, 0.13–0.93).

Exploratory analyses

Stratification of the analysis by parity revealed a higher proportion of gestational hypertension, pre-eclampsia, pre-term birth and low birthweight among primiparous women compared with multiparous women (Supplementary Tables S1–S4, available as Supplementary data at IJE online). Among primiparous women, periconceptional OC use was associated with pre-eclampsia (adjusted RR 1.44, 95% CI 0.99–2.10) and with fewer cases of gestational diabetes (0.65, 0.42–1.01) and SGA (0.71, 0.52–0.97). Periconceptional OC use among multiparous women was associated with SGA (1.41, 0.99–2.01).

Post hoc mediation analyses found that the effect of periconceptional OC use on pre-term birth was not mediated by pre-eclampsia (average proportion mediated 0.03, 95% CI –0.01 to 0.14). The effect of periconceptional OC use on low birthweight, however, was fully mediated by pre-term birth (0.96, –0.92 to 2.18).

Discussion

The aim of this study was to determine whether periconceptional OC use was associated with pregnancy complications and adverse birth outcomes. We observed associations between any periconceptional OC use and small to moderately increased risks of pre-eclampsia, pre-term birth and low birthweight. Our analyses of exposure characteristics revealed that these effects were largely attributable to OCs being used until 0–3 months before conception. Pre-term birth was the only outcome associated with OC discontinuation for ≤1 year before conception. Given that the association with low birthweight was fully mediated by pre-term birth, it is likely that an association between low birthweight and OC discontinuation 4–12 months prior to conception would also be found at larger sample sizes. No mediation effect of pre-eclampsia was found for periconceptional OC use with pre-term birth or, by extension, low birthweight. Furthermore, periconceptional use of OCs with low to moderate oestrogen doses (<30 µg) was moderately associated with pre-term birth, low birthweight and SGA. High-dose OC use (≥30 µg) was linked to slightly increased risks of pre-eclampsia and pre-term birth, and to a decreased risk of SGA. Using first- or second-generation OCs was weakly associated with pre-eclampsia and pre-term birth, whereas third-generation OCs were moderately associated with pre-term birth, low birthweight and SGA. Fourth-generation OCs were negatively associated with SGA, however. No associations were observed with gestational diabetes and gestational hypertension.

Whereas reported data appear to agree that using OCs increases blood pressure, which in turn may relate to the increased risk of gestational hypertension and pre-eclampsia, it remains unclear to what extent this depends on the oestrogen or progestin components.⁷ The risk of hypertension decreases with OC discontinuation, but the time frame of lasting effects remains unclear.⁷ To the best of our knowledge, no prior studies adequately investigated the associations between periconceptional OC use in the year before or during pregnancy and gestational hypertension or pre-eclampsia. Thadhani et al.⁸ found a negative association with gestational hypertension (adjusted RR 0.7, 95% CI 0.4–1.0) and a positive association with pre-eclampsia (RR 2.1, 95% CI 1.1–4.2, only among participants who had used OCs for ≥8 years), but their periconceptional period extended for ≤2 years prior to conception. Magnussen et al.¹² observed a negative association with pre-eclampsia (adjusted RR 0.5) but the average time between delivery and exposure assessment in this retrospective study was 3.5 years. In line with our findings on gestational hypertension, Farley et al.¹³ assessed contraceptive use at the time of conception and did not find an association with gestational hypertension (adjusted odds ratio 1.22, 95% CI 0.75–1.98) but they grouped OC users together with those using other non-barrier methods.

No prior studies exist that have investigated periconceptional OC use with gestational diabetes. Only Kramer et al.¹⁴ reported that any past usage (no time limit) of hormonal contraceptives (including OCs, injections, patches, cervical rings and intrauterine devices) increased the risk of gestational diabetes compared with no contraception. Given our findings and the review of evidence that OC use may only minimally dysregulate glucose control,⁴³ an association between OC use and gestational diabetes seems unlikely for those discontinuing before conception.

Our results on adverse birth outcomes generally agree with findings from previous studies with >500 exposed pregnancies.^9–11 Pardthaisong and Gray⁹ were the first to report a 50% increased risk of low birthweight after periconceptional OC use. Chen et al.¹⁰ investigated OC exposure in 1, 2 or 3 months before the last menstrual period and only observed increased risks of low birthweight or pre-term birth if used within 1 month of the last menstrual period. Jensen et al.¹¹ included the largest sample size to date [N = 44 834, of which 18 584 (41.5%) were exposed to periconceptional OC use] and found an association with pre-term birth across OC discontinuation times (12 months pre-conception to 12 weeks post-conception) but not with SGA. In the study by Hatch et al.,¹⁶ no association with SGA was observed (adjusted RR 0.91, 95% CI 0.73–1.13). Interestingly, their results indicated a decreased risk of low birthweight (0.69, 0.47–1.01). As their cohort consisted of a smaller proportion of first- and second-generation OC users (22.2% vs 28.7% in the current study) and a larger proportion of fourth-generation OC users (15.1% vs 3.4%), these results may align with our results of SGA for fourth-generation OCs.

Another study among 657 exposed pregnancies based on a cohort from 1974 did not find an association between periconceptional OC use and low birthweight (effect estimates not reported).¹⁵ However, this study may have been underpowered as the exposure group was divided into four OC discontinuation timings and by infant sex, leading to groups of exposure characteristics of ≤173 pregnancies for the analysis.

Observations from our analyses of exposure characteristics are broadly in line with the literature: oestrogen-containing OCs are more often associated with adverse effects compared with progestin-only pills.^7,43,44 One of the more notable findings is the increased risk of SGA with periconceptional OC use containing low to moderate oestrogen levels, whereas the effect seems to be protective with high-oestrogen-dose OCs. There is no consensus on the issue as both positive^45,46 and negative⁴⁷ associations between oestrogen level and birthweight were previously described. Periconceptional use of pills containing norethindrone (first-generation), levonorgestrel (second-generation) and desogestrel (third-generation) have also been associated with pre-term birth.¹¹ The finding that periconceptional use of fourth-generation OCs was associated with fewer cases of SGA may be attributed to a relatively small exposure groups (n = 201) and requires replication. Future studies of sufficient sample size should consider interaction terms between both oestrogen and progestin components.

The possible biological mechanism through which periconceptional OC use may affect pregnancy complications and birth outcomes remains to be elucidated. Effects may result from metabolic or hormonal changes persisting after OC discontinuation. Indeed, current OC use has been associated with lower serum hormone binding globulin levels and increased levels of C-reactive protein,^48,49 and some changes seem to persist for multiple months after discontinuation.^48,50

The most important strengths of this study are the use of far-reaching participant recruitment methods, prospective data collection via detailed questionnaires at three time points during pregnancy, the possibility to complete the questionnaires online and access to the majority of obstetric records to account for missing or inaccurate outcome data. Data on OC use were also self-reported, which is more reliable than pharmacy records during the periconceptional period.^51,52 Another strength was the sophisticated adjustment for confounding and potential selection bias due to missing outcome data.

The main limitations of the study are the small sample sizes and number of events for some exposure characteristics of interest. Generally, we consider our findings related to pregnancies exposed to OCs after conception or took progestin-only or fourth-generation OCs to be underpowered. However, this reflects real-world use: it will always remain rare for women to be using OCs after conception. Similarly, progestin-only OCs and those classified as first- or fourth-generation are unpopular. Another limitation may be that most eligible participants were Dutch (91.0%) and relatively highly educated (73.6%). However, previous studies did not foresee issues in external validity despite the selection process.^53,54 Finally, despite the use of propensity scores and adjustment, residual and unmeasured confounding may still exist from other variables not considered in the study, e.g. diet, exercise, marital status, household income and obstetric history.

In conclusion, periconceptional OC use was weakly to moderately associated with pre-eclampsia, pre-term birth and low birthweight (the latter being fully mediated by pre-term birth). Variable associations were observed for SGA with different amounts of oestrogen and with pills containing a progestin component of the third or fourth generation. No associations were found with gestational diabetes and gestational hypertension. Associations with pre-eclampsia were found for OCs discontinued within 3 months before pregnancy, with a high oestrogen content and those containing a first- or second-generation progestin. Periconceptional use of OCs with low to moderate oestrogen amounts or third-generation progestins, however, increased the risk of adverse birth outcomes of interest considerably. Therefore, women who intend to become pregnant within the next year might benefit from initiating or switching to a non-hormonal option. However, this should be balanced against the substantial public health benefits of OCs, including prevention of mistimed or unintended pregnancies, treatment of menstrual disorders and associated anaemia, and prevention of ovarian cysts, endometrial and ovarian cancers, and osteoporotic fractures. More international confirmatory studies are required to support our findings.

Ethics approval

This study was performed in line with the principles of the Declaration of Helsinki. Approval was granted by the Regional Committee on Research Involving Human Subjects Arnhem-Nijmegen (CMO 2009/305).

Data availability

The data underlying this article cannot be shared publicly due to privacy and ethical restrictions. The data will be shared on reasonable request to the corresponding author.

Supplementary data

Supplementary data are available at IJE online.

Author contributions

All authors contributed to the study conception and design. Data collection and cleaning were performed by I.M., N.S. and M.M.H.J.v.G. The analysis was performed by A.S., A.L. and M.M.H.J.v.G. The first draft of the manuscript was written by A.S. and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

Funding

None.

Conflict of interest

None declared.