Phenome‐Wide Analysis of Short‐ and Long‐Run Disease Incidence Following Recurrent Pregnancy Loss Using Data From a 39‐Year Period

Abstract

Background

It is unclear how recurrent pregnancy loss (RPL) impacts disease risk and whether there is a difference in risk between women with or without a live birth before RPL (primary versus secondary RPL). We investigated the disease risk following RPL, and whether there was a difference between primary and secondary RPL.

Methods and Results

Using population‐wide healthcare registries from Denmark, we identified a cohort of 1 370 896 ever‐pregnant women aged 12 to 40 years between 1977 and 2016. Of this cohort, 10 691 (0.77%) fulfilled the criteria for RPL (50.0% primary RPL). Average follow‐up was 15.8 years. Incidence rate ratios were calculated in a phenome‐wide manner. Diagnoses related to assessment and diagnosis of RPL and those appearing later in life were separated using a mixture model. Primary RPL increased the risk of subsequent cardiovascular disorders, including atherosclerosis, cerebral infarction, heart failure, and pulmonary embolism, as well as systemic lupus erythematosus, chronic obstructive pulmonary disease, anxiety, and obsessive‐compulsive disorder. Women with secondary RPL had no increased risk of cardiovascular disorders. However, we observed an increased risk of gastrointestinal disorders such as irritable bowel syndrome and intestinal malabsorption, as well as mental disorders and obstetric complications.

Conclusions

RPL is a risk factor for a spectrum of disorders, which is different for primary and secondary RPL. Screening following RPL explains some associations, but the remaining findings suggest that RPL influences or shares cause with cardiovascular disorders, autoimmune disorders, and mental disorders. Research into the pathophysiology of RPL and later diseases merits further investigation.

Nonstandard Abbreviations and Acronyms

PL

pregnancy loss

RPL

recurrent pregnancy loss

ICD‐10

International Classification of Diseases, Tenth Revision

Clinical Perspective

What Is New?

• Recurrent pregnancy loss (RPL) is associated with disease later in life; understanding the link between RPL and subsequent disease patterns is crucial to identify prevention and treatment.

• Women experiencing primary and secondary RPL have different subsequent disease patterns, notably within the cardiovascular, autoimmune, mental, and endocrine disease spectrum.

• Only primary RPL is associated with later cardiovascular disease.

What Are the Clinical Implications?

• Primary and secondary RPL should be considered as 2 separate risk factors for later disease.

• Clinicians should include pregnancy history (including information about miscarriages) when assessing disease risk in a clinical setting and in study designs.

• Primary RPL is especially important for the assessment of cardiovascular disease risk.

Introduction

Recurrent pregnancy loss (RPL) is a condition that affects 1% to 3% of couples, depending on its precise definition.1, 2 While the definition of RPL is not fully consistent across the world, it has historically been defined as 3 or more consecutive pregnancy losses (PL), and this is the definition used in Denmark and this study.3 With increasing number of losses, the frequency of euploid losses increases while the chance of a live birth decreases.4, 5, 6 RPL is divided into 2 categories: primary (no live birth before losses) and secondary RPL (at least 1 live birth before losses). Prior Nordic studies report that 40% to 48% of cases are secondary RPL, but in some countries this may be as high as 61%.1, 7, 8

RPL is a multifactorial condition that, in addition to fetal causes, have multiple known female risk factors such as autoimmune diseases, endocrine disorders, and uterine malformations.2, 9 Risk factors differ between women with primary and secondary RPL. Male risk factors have been investigated less, but include chromosomal abnormalities, sperm DNA fragmentation, and age.10

In recent years, PLs and RPL have been suggested or identified as a risk factor for cardiovascular disease (CVD) and malignancies, although the evidence is ambiguous for the latter.11, 12, 13, 14, 15, 16 The association with CVD has been further explored in a study that found first‐degree relatives of women with RPL had an increased risk of CVD.17, 18 It has yet to be investigated if the risk is different between women with primary and secondary RPL. Some factors associated with a delayed time to pregnancy and infertility, such as endometriosis, polycystic ovary syndrome, and diabetes mellitus, have a clear association to disease later in life, including malignancies, autoimmune diseases, and CVD.19, 20, 21 Moreover, obstetric complications have also been associated with an increased risk of CVD.22 Despite RPL having a major impact on the affected couples, both mentally and physically, the cause and lifelong consequences are poorly understood.23, 24 Our hypothesis is that RPL and PLs not attributed to aneuploidy are associated with later disease, and that women with primary and secondary RPL have quite different risk profiles in terms of later disease. Identifying novel associations could lead to a pathophysiological understanding that can improve the take‐home baby rate, long‐term maternal health, and reduce the health gap between men and women.

In this study, we compare primary and secondary RPL in a phenome‐wide analysis using a nationwide cohort based on 1 370 896 women.

Methods

The study was approved by the Danish Data Protection Agency (ref: 2015‐54‐0939 and SUND‐2017‐57) and Danish Health Authority (ref: FSEID‐00001627 and FSEID‐00003092). Informed consent and assessment of the proposal in scientific ethical committees are not required for registry‐based research in Denmark. Permission to access and analyze data can be obtained following approval from the Danish Data Protection Agency and the Danish Health Authority.

Study Design and Setting

This observational population was identified using the nationwide Danish Patient Registry and the Danish Medical Birth Registry. We included women aged 12 to 40 years with at least 1 live birth or PL in the period between January 1, 1977 and October 5, 2016. Women with multiple births (twins or higher order) were excluded. The population totaled 1 370 896 women.

Women with RPL were followed from the date of meeting the exposure until the end of follow‐up (criteria until death or October 5, 2016).

Each woman with RPL was matched with 20 women from the population without RPL. The women were matched based on year of birth and number of previous live births. The matched comparison group of women without RPL was followed for the outcomes from the same date of the woman with RPL they were matched to (ie, only the hospital admissions occurring on or after the date of RPL) and matching was used in the analysis. Women with RPL were divided into primary and secondary RPL based on the parity history before the date of RPL.

Outcomes were identified using several nationwide registries that cover all hospital admissions in Denmark (see Data S1 for a description).25

We identified 10 691 women who fulfilled the criteria for RPL (50.0% primary) (defined in the next section). The number of years the women were followed varied by outcome, but the average number of person‐years was 1 782 238 years (range 1 231 217–1 799 076 years) for women with primary RPL and the matched group, and 1 744 736 years (range 1 155 427–1 763 733 years) for women with secondary RPL and the matched group. Across all outcomes, the median follow‐up time for women was 15.8 years, ranging from 11.6 to 16.4 years for each specific outcome.

Pregnancy Loss and Recurrent Pregnancy Loss

PLs were identified from hospital admissions in the Danish National Patient Registry (Table). PLs occurring 8 weeks before or after a molar pregnancy, induced abortion, or extrauterine pregnancy were disregarded (Table). In Denmark, every child showing signs of life at delivery is categorized as a live birth irrespective of gestational age. If there is no sign of life at delivery it is considered a PL before the 28th completed weeks of gestation and a stillbirth thereafter. This definition was changed starting from 2004 when stillbirth was counted from 22nd completed weeks of gestation. We filtered out diagnoses that were repeated within a medically unreasonable period of time: (1) <22 weeks between 2 live or still births, (2) <90 days between 2 PLs, termination of pregnancy, molar or ectopic pregnancies, (3) <22 weeks between a PL, induced abortion, or an extrauterine or molar pregnancy and a stillbirth or livebirth, and (4) <30 days between a live or still birth and a PL, termination of pregnancy, or an extrauterine or molar pregnancy.

Table 1.

ICD‐8 and ICD‐10 Codes Used to Identify Recurrent Pregnancy Loss, Pregnancy Loss, and Molar and Extrauterine Pregnancies

Diagnosis	ICD‐8/ICD‐10	Surgery	Procedures
Recurrent pregnancy loss	6430x, Y6439, N96.x, O26.2	···	···
Missed abortion	6346x, 6451x, O02.1, O02.1A	···	···
Miscarriage	6438x, 6439x, O03.x	···	···
Extrauterine pregnancy	63109, 6311x, 63129, 63139, 63149, 6315x, 63169, 63199, O00.x	66100, KLBCx, KMCBx,	BKHE0, BKHE8x
Molar pregnancy	63190, 63429, 63460, 6450x, D39.2, O01.x, O02.0,
Abortion (induced)	6400x, 6401x, 6402x, 6409x, 6410x, 6411x, 6412x, 6413x, 6414x, 6415x, 6416x, 6417x, 6419x, 64209, 64219, 64229, 64239, 64299, 6455x, 6456x, 6458x, O04.x, O05.x, O06.x, O07.x	63680, 94520, KLCHx, KLWW00,	BKKG1, BKXG1

Lowercase x denotes codes including all subcodes. ICD‐8 and ICD‐10 indicate International Classification of Diseases, Eighth and Tenth Revisions.

Cases of RPL were identified using hospital discharge codes (Table) or by 3 consecutive PLs. The date of RPL diagnosis was defined as the date of the third PL or RPL diagnosis, whichever came first.

In the matching process, we used 1 370 339 women from the total population. The median number of unique matched women per outcome was 186 752, ranging from 133 875 to 187 355.

Outcomes

In the phenome‐wide analysis, we investigated International Classification of Diseases, Tenth Revision (ICD‐10) codes at the third level that had a prevalence of ≥0.1% in women with RPL (626 diagnoses). ICD‐10 codes related to birth, PL, and termination of pregnancy were excluded (O00–O08 and O80–O84), as well as codes from ICD‐10 chapters 19, 20, 21, and 22. Only the first instance of a diagnosis code was included.

Differentiating Early From Late Disease Occurrences

Some of the diagnoses with an increased risk are because of investigations at an RPL clinic. To differentiate the diagnoses found during investigations at RPL clinics and later in life, we fitted a log‐normal mixture model. The optimal number of components was determined using the Bayesian Information Criteria. For each diagnosis code with an increased risk, the time from RPL diagnosis until the outcome occurred was summarized as the median. Time distributions were visualized as histograms that binned multiple diagnoses into 1 bin.

Statistical Analysis

Incidence rate ratios (IRR) were estimated using a hierarchical Bayesian Poisson log‐linear model with time in 3‐year bands. As covariates we included the age, binned into 11 groups (Table S1), and previous number of live births. A model was fit separately for women with primary and secondary RPL, respectively. Bayesian posterior distributions were summarized as 95% uncertainty intervals (UI). To account for multiple testing, we calculated the false discovery rate using the Benjamini and Hochberg method. An association was deemed significant if the false discovery rate was <5%. We defined a prior structure that takes into account the chapter structure in the ICD‐10 terminology (see Data S1 and Figure S1 for a thorough evaluation of the priors and the model). Distributions of time from RPL diagnoses until outcome were modeled using a log‐normal mixture model, and the number of components was selected using the Bayesian Information Criteria (Table S2). Additional details on the statistical procedures are provided in Data S1. All data were analyzed using stan v2.18, python v2.7, and R v3.1.3.

Results

Out of a pool of 1 310 unique ICD‐10 codes assigned to women after RPL, we investigated 615 diagnoses with a cumulative incidence proportion of at least 0.1% in women with RPL. We found 151 and 127 diagnoses with an elevated IRR in primary and secondary RPL, respectively. These covered very heterogeneous types of disease (Figure 1, Data S1), and included CVDs, obstetric diagnoses, autoimmune diseases, mental disorders, gastrointestinal disorders, and respiratory diseases. We also observed 7 diagnoses with a RR lower than 1 (all in women with primary RPL), which were all related to complications of labor and delivery (Data S1).

Figure 1. Diagnoses occurring more frequently following recurrent pregnancy loss (RPL) across 18 ICD‐10 chapters.

The points have been scattered in the vertical direction to improve readability. Each point represents the median value from the posterior distribution. A, Diagnoses with an elevated risk for women with primary RPL (180 diagnoses). B, Diagnoses with an elevated risk for women with secondary RPL (172 diagnoses). ICD‐10 indicates International Classification of Diseases, Tenth Revision.

We identified an “early” and “late” component for both primary and secondary RPL (Figure 2, Data S1). The median time to each diagnosis for women with RPL is available in Data S1. In women with secondary RPL, the “late” component had a mean value of 11.06 (SD=3.16) years after the diagnosis. Diagnoses that only had an increased risk in women with secondary RPL included irritable bowel syndrome (IRR=1.32, 1.10–1.60 95% UI) and intestinal malabsorption (IRR=1.67, 1.11–2.45 95% UI). Mental disorders included recurrent depressive disorders (IRR=1.35, 1.17–1.55 95% UI) and mixed personality disorders (IRR=1.58, 1.11–2.24 95% UI).

Figure 2. Distribution of the median time between a recurrent pregnancy loss (RPL) diagnosis and 1 of the 180 and 172 diagnoses that are more frequent following primary RPL and secondary RPL, respectively.

The histogram indicates the observed data points, whereas the 2 density plots indicate the 2 components of the mixture model.

The late component was found to have a mean of 12.6 (SD=4.53) after the primary PRL diagnosis and contained 141 diagnoses. Strikingly, for many CVDs we did not observe an increased risk in women with secondary RPL (Figure 3). This included atherosclerosis (IRR=1.90, 1.30–2.75 95% UI), cerebral infarction (IRR=1.83, 1.36–2.38 95% UI), heart failure (IRR=1.93, 1.40–2.63 95% UI), and pulmonary embolism (IRR=1.64, 1.16–2.29 95% UI). Other diseases included lupus erythematosus (IRR=3.79, 2.45–5.78 95% UI), systemic lupus erythematosus (IRR=2.52, 1.34–4.31 95% UI), chronic obstructive pulmonary disease (IRR=1.56, 1.24–1.96 95% UI), and type 2 diabetes mellitus (IRR=1.92, 1.62–2.27 95% UI). Several mental disorders were also found, including anxiety (IRR=1.29, 1.08–1.53 95% UI) and substance abuse because of smoking (IRR=2.0, 1.65–2.42 95% UI).

Figure 3. Diagnoses unique to either primary or secondary recurrent pregnancy loss, divided into the 2 “early” and “late” groups determined from the mixture model analysis.

Each point represents the median value from the posterior distribution. The coloring scheme is as in Figure 1. ICD‐10 indicates International Classification of Diseases, Tenth Revision.

Several disorders were found to have an increased risk in both subtypes, such as thyroid disorders (hypothyroidism and thyroiditis), asthma, reaction to severe stress, depressive episodes, risk of mental disorders because of substance abuse of alcohol, and gastrointestinal disorders (gastroesophageal reflux disease, and gastritis and duodenitis). The disorders all belonged to the late component.

The early component was found to have a mean of 2.26 (SD=0.78) years for primary RPL and 1.68 (SD=0.56) years for secondary RPL. The component contained diagnoses routinely investigated as part of the RPL evaluation (eg, uterine malformations, balanced chromosomal rearrangements, and abnormal immunological findings in serum). Cervical malformations, which is not a part of the standard evaluation, was also found to have an increased risk. Additionally, there were many diagnoses related to obstetric complications and pregnancy for both primary and secondary RPL. Certain obstetric complications were only found to have an increased risk in secondary RPL, such as placenta previa (IRR=3.67, 95% UI 2.80–4.77), premature rupture of membrane (IRR=2.67, 95% UI 2.30–3.06), intrapartum hemorrhage (IRR=2.53, 95% UI 1.55–3.98), gestational hypertension (IRR=1.96, 95% UI 1.45–2.63), pre‐eclampsia (IRR=2.29, 95% UI 1.81–2.87), puerperal sepsis (IRR=2.55, 95% UI 1.80–3.54), and placental abruption (IRR=2.91, 95% UI 2.15–3.86).The latter also had an increased risk in primary RPL, albeit lower (IRR=1.32, 95% UI 1.05–1.67).

Discussion

In this study, we performed the largest registry‐based phenome‐wide study to date of longitudinal disease incidence in RPL. The population comprised 1 370 896 women, of which 10 691 had 3 or more consecutive PLs. By investigating diagnoses occurring after RPL, we identified distinct spectrums of complications for primary and secondary RPL. The diagnoses were divided into “early” and “late” complications: The “late” component included multiple complications in heterogeneous disease domains, such as CVDs, autoimmune diseases, mental disorders, digestive system disorders, and respiratory diseases. The disorders occurred, on average, >10 years after the RPL diagnosis, and there was a distinct set of diseases associated with primary and secondary RPL. The “early” complications contained many previously established risk factors for RPL that are part of the routine screening at RPL clinics.2 These included coagulation disorders, congenital malformations of the female genital organs, and autoimmune diseases. The early component also contained obstetric complications, with the majority seen only after secondary RPL. We found no evidence that women with RPL had an increased risk of malignancies, irrespective of the subtype. The large population of women, both with and without RPL, allowed us to investigate late‐state complications in a phenome‐wide manner. This has historically not been possible, because of the small cohort sizes often used. The findings we present here are of wide interest as we identify primary and secondary RPL as a risk factor and an early indicator of later disease.

The basis for this study was nationwide data collected over a 39‐year period. Reporting to the Danish registries used in this study is mandatory. An important source of bias is unregistered PLs. This includes PLs handled at home, in general or gynecological practice. This leads to some women being erroneously classified as not having RPL, leading to an underestimation of the IRR. Furthermore, the RPL definition depends on the reliability of the identification of PLs in the Danish National Patient Registry. However, this condition has previously been estimated to have a positive predictive value of 97%.26 Lack of information from the Danish Medical Birth Register could result in women being classified wrongly as primary and secondary RPL; nonetheless, the Danish Medical Birth Register is considered to be complete.27 The registries do not contain complete information on smoking status or body mass index. Additionally, we did not have any information on socioeconomic status or whether the partner was the same for all pregnancies, which could confound our findings. Furthermore, we did not include the disease history before RPL, which could also confound the results. Still, many women experiencing RPL are not properly investigated before referral to a specialist clinic, and it is therefore not possible to take this into account. We also note that the infant and maternal mortality in Denmark is extremely low.28, 29 This is partially because of universal health care and referral to specialist clinics involved in the monitoring and treatment of, for example pregnant women with diabetes mellitus and heart disease. Therefore, we would not expect to see many PLs in the cohort because of, for example, uncontrolled diabetes mellitus. Additionally, there are no known causal factors for RPL with the exception of embryonal malformations, and it is thus difficult to control for pre‐existing conditions related to RPL. The probability of an early PL is highly correlated with age. Since we did not have information on the karyotype of the product of conception, we have tried to minimize the confounding effect from aneuploidy by only including women younger than 40 years of age. Moreover, because of the lack of paternal information, we could not account for those cases where there is a paternal chromosomal translocation. A mixture of cases because of aneuploidy or paternal translocations would weaken the signal, and we could thereby have missed some associations, or underestimated the relative risks. This study was based on population data from a 39‐year period. While this is helpful for a long follow‐up period, changes and developments in medical practice may influence the results. However, we did attempt to mitigate this effect by matching women born in the same year. Lastly, because of the large number of outcomes studied in this phenome‐wide analysis, we cannot rule out chance findings. Nevertheless, we have attempted to mitigate this by advanced statistical modeling and we also note that the large number of diagnoses associated with RPL remains striking and is consistent within certain domains of disease, which merits further investigation.

We found that primary and secondary RPL, when considered separate phenotypes, are followed by a unique spectrum of complications later in life. Our observations in the domain of CVDs and obstetric complications are in accordance with prior studies.12, 13, 30, 31 Yet, previous studies of CVD and PL have not been stratified by subtype, and here we demonstrate that primary RPL drives the observed associations. This is an important distinction, because this could point towards shared pathophysiology only present in women with primary RPL and could serve as the basis for future screening and risk‐assessment profiles. Furthermore, we found that systemic lupus erythematosus (SLE), a recognized risk factor for RPL, also was significantly more frequent in women after an RPL diagnosis and belonged to the “late” component. This can be explained in at least 2 different ways. First, if SLE and RPL share a common pathophysiological cause, our findings indicate that this cause has not always manifested fully at the time of RPL diagnosis and evaluation or it could be that RPL itself increases the risk of SLE, possibly through increased microchimerism.32, 33 Lastly, the increased risk of mental disorders and substance abuse could possibly be mitigated by a closer follow‐up and referral to therapy.

There are large discrepancies across health in men and women.34 Here, we have tried to elucidate how the pregnancy history contributes to disease. The identification of complications across the full spectrum of disease present a new set of challenges that must be examined in‐depth clinically to uncover the cause and close the gap in health care between men and women. Identification of clinically relevant subtypes is an important aspect of precision medicine, both to tailor screening and therapy to reduce the disease burden, but also to prevent overdiagnosis and unnecessary invasive procedures in an already vulnerable patient group. We have demonstrated this aspect in our distinction between primary and secondary RPL, in which CVD only had an increased risk in 1 subtype. Indeed, RPL may very well be an early initiator of disease, disease maintenance, and disease progression. However, the exact relationship between RPL and chronic diseases remains to be elucidated. Most importantly, our results indicate that RPL is an early marker of a wide range of diseases. We therefore speculate that, in the future, fertility history will be a core part of most risk prediction models across diverse clinics in health care. Considering the full pregnancy history including prior PLs is thus relevant when evaluating and predicting the future risk of disease in women.

Sources of Funding

The work is carried out as a part of the BRIDGE – Translational Excellence Programme (bridge.ku.dk) at the Faculty of Health and Medical Sciences, University of Copenhagen, funded by the Novo Nordisk Foundation (grant agreement NNF18SA0034956). Funding from other Novo Nordisk Foundation grants (NNF14CC0001 and NNF17OC0027594) and the Ole Kirk Foundation and Rigshospitalet's Research Fund is also acknowledged.

Disclosures

None.

Supporting information

Data S1

Tables S1–S2

Figure S1

References 35–47

(J Am Heart Assoc. 2020;9:e015069 DOI: 10.1161/JAHA.119.015069.)