Background and objective Previous studies suggest that women with coeliac disease (CD) have reproductive difficulties but the data are often inconclusive and contradictory. Fertility in women with biopsy-verified CD was examined.
Methods Swedish population-based cohort study. Duodenal/jejunal biopsy data on CD (Marsh III; villous atrophy (VA); n=18 005 unique women) were collected from all (n=28) pathology departments in Sweden. From this dataset, 11 495 women with CD, aged 18–45 years, were identified at some point before the end of follow-up. Multinomial logistic regression and Cox regression were used to estimate fertility in these women compared with that in 51 109 age-matched reference women. Fertility was defined as the number of children according to the Swedish Multi-Generation Register.
Results During follow-up, 16 309 births occurred in women with CD and 69 245 in the reference women. The cumulative number of children slightly increased in women with CD compared with the reference group. Adjusting for age, calendar period and parity and stratifying by education, the overall fertility hazard ratio (HR) in CD was 1.03 (95% CI 1.01 to 1.05). Specifically, the fertility HR was 1.05 (95% CI 0.96 to 1.14) for CD diagnosed in women before 18 years, 1.04 (95% CI 1.01 to 1.07) for CD diagnosed in women between 18 and 45 years and 1.02 (95% CI 0.99 to 1.04) for CD diagnosed in women > 45 years of age. Taking date of CD diagnosis into account, fertility was decreased 0–2 years before time of diagnosis (HR=0.63; 95% CI 0.57 to 0.70), was identical to that of controls 0–5 years subsequent to diagnosis and increased to 1.12 (95% CI 1.03 to 1.21) thereafter.
Conclusion Overall, women with CD had a normal fertility, but their fertility was decreased in the last 2 years preceding CD diagnosis.
- celiac disease
Statistics from Altmetric.com
Significance of this study
What is already known about this subject?
Patients with coeliac disease are at increased risk of adverse pregnancy outcome.
Women with coeliac disease may have a shorter fertile timespan because of later menarche and earlier menopause.
Earlier studies on fertility and infertility in coeliac disease are contradictory.
What are the new findings?
Overall, coeliac disease was not associated with decreased fertility.
Fertility in coeliac disease was, however, decreased in the last 2 years preceding diagnosis.
Age at diagnosis did not influence fertility in women with coeliac disease.
How might they impact on clinical practice in the foreseeable future?
Women with coeliac disease should be informed that active coeliac disease may decrease fertility, but that treatment with a gluten-free diet is likely to normalise fertility.
Coeliac disease (CD) is an immune-mediated disorder affecting some 1% of the Western population,1 with recent data suggesting a prevalence of 1.6% among Swedish adults.2 In recent years it has become evident that many patients with CD develop extraintestinal disorders and symptoms, including malignancy,3 fractures4 and adverse pregnancy outcome. Offspring to women with undiagnosed CD are at increased risk of both preterm birth and low birth weight.5 6
Although it is usually suggested that CD is associated with infertility,1 7 conclusive evidence remains elusive. Earlier research is contradictory,8–19 and the prevalence of CD among women with infertility varies: 0.8%,19 2.1%,11 2.5%,20 2.6%,13 3.0%,10 5.0–7.5%21 (unclear if published data21 refer to biopsy-verified patients or only antibody-positive individuals).
Furthermore, several studies have not included control groups. In a Finnish study13 the statistically significant association between CD and unexplained infertility (p=0.02; although a weaker relationship was found to any infertility: p=0.06) could be partly explained by the unexpectedly low prevalence of CD in the healthy control group (0/150). Using published data from the study of Meloni et al,10 we recalculated their risk estimate and found no association between CD and any infertility (Fisher's exact test). Several other studies also suggest that CD is not linked to infertility.18–20 The only available cohort study on fertility found a similar fertility in CD and controls (fertility rate ratio in CD=1.01).18 Nevertheless, reproductive organs seem to be affected in women with CD22 and women with CD may have a shorter fertile timespan because of later menarche and earlier menopause. Moreover, folic acid deficiency is often seen in CD (at diagnosis23 and later24), which could decrease fertility.25
In this study we collected data from biopsy reports in all (n=28) pathology departments in Sweden, allowing us to identify more than 11 000 women in fertile age and with biopsy-verified CD.26 Linkage with the Swedish Multi-Generation Register (MGR) then allowed us to identify all offspring and thus to compare their fertility rates with randomly selected reference individuals.
We used Swedish regional biopsy data to identify patients with CD.26 Between 2006 and 2008, we collected pathology reports from biopsies of the small intestine that were carried out between 1969 and 2008 at any of the 28 pathology departments in Sweden. For each biopsy, we recorded arrival date of biopsy, unique 10-digit personal identity number, morphology and topography (duodenum or jejunum). We used SnoMed morphology classification codes equivalent to villous atrophy (VA) (Marsh III27) to define CD (see online supplementary table). Validation against patient charts has shown that 95% of individuals with VA have CD.26 After the exclusion of duplicates and data irregularities (eg, recorded date of biopsy before birth or after death), 351 403 biopsy reports remained for 287 386 unique individuals (110 479 men and 176 907 women). This study is restricted to 18 005 women who had a record of duodenal/jejunal VA.
For each woman with CD, the government agency Statistics Sweden identified up to five reference individuals matched by age, sex, calendar period and county of residence. Individuals diagnosed with CD during the follow-up and those who were not alive or were not in Sweden at the time of the diagnosis of CD of the corresponding case were excluded.
We used the Swedish MGR to identify all children born to women with CD and their reference individuals. The Swedish MGR contains information on the parents of all individuals in Sweden born from 1932 onwards and who were alive in 1961.28 Adoption or other non-biological relations are flagged in the register, with non-biological and half-siblings excluded from all analyses. Linkage with the MGR was made possible through the personal identity number, which is assigned to more than 99% of all Swedish residents.29 Migration dates were obtained from the Migration Register, a register that contains data since 1960.
Through linkage with the Education Register, we obtained information on the number of years of formal education completed as of 2005. Women diagnosed before the age of 18 years were assigned the education level of their fathers or the education level of their mothers if paternal education was missing. From the Total Population Registry, we obtained data on the country of birth of each woman.
We restricted the study to women born between 1920 and 1990 whose reproductive history was covered by the MGR starting from age 18 years, leaving 12 875 patients with CD and 64 118 reference individuals in the study. Women who were not present in Sweden at 18 years of age (705 patients with CD and 4569 reference individuals) were excluded. We also excluded all women who gave birth before reaching age 18 (277 patients with CD and 1525 reference individuals), those born outside the Nordic countries or whose origin was unknown (182 patients with CD and 1776 reference individuals) and those with no data on educational level (216 patients with CD and 886 reference individuals). When a patient with CD was excluded from the study, we also excluded the corresponding matched women. The main analysis was based on 11 495 women with CD and 51 109 reference women.
This study was approved by the research ethics committee of the Karolinska Institutet, Stockholm, Sweden.
Women were followed up from 18 years of age to the age of 44, emigration, death or 1 January 2008, whichever occurred first. The three CD groups were defined a priori according to the age at the time of diagnosis: before reproductive age (<18 years), during reproductive age (between 18 and 44 years) and after reproductive age (≥45 years). In all analyses we maintained the matching design by comparing each CD group with the corresponding matched reference group.30
We used multinomial logistic regression31 to evaluate the association between CD groups and the cumulative number of children during the complete fertile period (18–44 years). To remove the effect of potential biases due to delayed pregnancy and catch-up phenomena this analysis was limited to 5912 CD women and 25 494 reference women with complete follow-up between 18 and 44 years of age. ORs were adjusted for women's year of birth categorised into 5-year classes and education level (categorised as listed in table 1).
To compare fertility hazards between patients with CD and reference women we used a Cox regression model for repeated events,32 33 with age as temporal axis. The model was stratified by education level to satisfy the proportionality assumption (checked by scaled Schoenfeld residuals). We included number of children and 5-year calendar periods as time-dependent variables in the model.
To investigate whether the actual diagnosis of CD (and its subsequent treatment) could affect the fertility HR we included a time-dependent variable for diagnosis of CD in the Cox model. Specifically, we evaluated different time intervals before (>5 years, between 2 and 5 years and <2 years) and after (<2 years, between 2 and 5 years and at least 5 years) CD diagnosis. This analysis was restricted to women diagnosed with CD during their reproductive age (ie, between 18 and 44 years). Data on date of conception were estimated by subtracting 40 weeks from the date of birth of the children.
All analyses were conducted using Stata release 9.0 (Stata Corp, College Station, Texas, USA).
In this cohort the majority of women with CD were diagnosed in adulthood, with 22% diagnosed before 18 years, 39% between 18 and 44 years and 39% after 45 years of age (table 1). We found no major differences in the distribution of education level and number of children between patients with CD and reference women.
In the analysis of cumulative number of children during fertile years we found a slightly increased number of children in women diagnosed with CD (table 2). None of the women diagnosed before 18 years of age had a complete follow-up until age 44 and these women were therefore excluded from this analysis on the total number of children.
Fertility HRs were slightly increased in all three groups of patients with CD (table 3). The HR of having a child for all patients with CD combined was 1.03 (95% CI 1.01 to 1.05).
Table 4 reports the HRs of supposed conception estimated at set intervals of time around biopsy date in women diagnosed with CD during their reproductive age. We found an increased fertility until 2 years before the diagnosis of CD. From that time on, the HR decreased to 0.63 (95% CI 0.57 to 0.70) in the 2 years before the diagnosis, remained close to one in the first 5 years subsequent to diagnosis and then increased to 1.12 (95% CI 1.03 to 1.21) thereafter.
This is the largest study on fertility in women with CD to date. Overall, women with CD had a normal fertility, but fertility decreased in the last 2 years preceding CD diagnosis. This decrease probably corresponds to the time of most active disease.
With the exception of a British study,18 most studies on fertility in CD have been restricted to either case–control studies10 11 13 14 19–21 34 or case series/case reports.16 35 36 In several studies patients originated from one or only a few centres10 11 13 16 19–21 34 and therefore these studies have a potential risk of selection bias.
In contrast, our cohort study consisted of women with CD identified from biopsy records that included all Swedish pathology departments. This population-based and national approach minimises selection bias. Considering that more than 96% of Swedish gastroenterologists and paediatricians perform a biopsy before CD diagnosis,26 collecting biopsy reports is a highly sensitive method for ascertaining individuals with a diagnosis of CD. The method is also specific with regards to CD. Among 114 randomly selected patients with VA, 108 (95%) had CD.26 This is actually higher than the positive predictive value of a CD diagnosis assigned by doctors in the Swedish national inpatient register (66/77, 86%).37
Using biopsy data also has the advantage of establishing an exact date of diagnosis. Inpatient (and some outpatient) registers sometimes suffer from a time lag in that some patients receive their diagnosis only months or years later. This fact has important implications when differentiating between undiagnosed and diagnosed CD, as gluten-free diet is usually initiated just after biopsy. Another strength of this study is the large number of participants (superseding that of all earlier studies of fertility and infertility in CD), which allowed us to calculate precise estimates of fertility rates and relative risks.
Earlier literature on reproduction in patients with CD
Earlier data on reproduction in patients with CD have been contradictory.8–19 In 2000, Gasbarrini et al failed to show an association between undiagnosed CD and recurrent spontaneous abortion for CD (3/40 patients with abortions were positive for endomysial antibodies versus 0/50 healthy blood donors; Fisher's exact test, p=0.084).21 In a subanalysis Meloni et al10 found an increased risk of CD in women with unexplained infertility (2 out of 25 women; p=0.033) but this association was not seen in women with any infertility (3 out of 99 women).
Our overall data are consistent with those of Tata et al,18 except for the decrease in fertility just before diagnosis. Another difference was that we found no relationship between age at CD diagnosis and fertility. One explanation for this discrepancy lies in the number of patients. The study of Tata et al was based on 341 births18 as compared with 16 309 in this study. Another explanation pertains to country-specific factors. Whereas their study suggested that a shift in fertility pattern over age might be due to socioeconomic status, adjustment for education in our study did not influence the results.
Interpretation of findings
Our data strongly indicate that infertility is more common in patients with active CD when a gluten-free diet is unlikely to have been initiated. However, in a life-time perspective, women with biopsy-verified CD seem to have similar fertility rates to those of women in the general population. We did not observe a significant difference in distribution of the cumulative number of children up to the age of 45 years, which we defined as the end of the fertile period.
Several circumstances could contribute to a lower fertility around the time of diagnosis of CD. We believe that high disease activity has a negative effect on fertility. In our earlier paper5 we found that the risk of adverse pregnancy outcome was highest just around the time of diagnosis (both before and after). Increased symptoms leading up to a diagnosis of CD could also diminish the desire to have a child. Furthermore, women undergoing investigation for CD might choose not to become pregnant until after the diagnostic investigation has been performed. This investigation sometimes takes more than a year because a large proportion of women has a control biopsy 1–2 years after the initial biopsy. Finally, a diagnosis of CD involves starting on a new (gluten-free) diet. Awareness of the need for optimal nutrition during pregnancy may cause some mothers to postpone pregnancy until they feel comfortable with the gluten-free diet.
One limitation of this study is the lack of data on smoking. Evidence indicates that smoking has a negative effect on fertility38 and if smokers were under-represented among patients with CD,39 this could lead to an overestimation of the fertility in CD. An earlier nationwide Swedish study from 2005, however, showed that smoking is not associated with undiagnosed CD in pregnant women,40 whereas it is positively associated with diagnosed CD. Hence, in a Swedish setting smoking is unlikely to explain our findings. We lacked data on body mass index. Patients with low body mass index are more likely to be investigated for CD and ultimately have a CD diagnosis.41 However, low body mass index is not associated with increased fertility42 and therefore should not explain our findings.
Since we only had data on diagnosed CD (through biopsy reports) we were unable to examine infertility among individuals who never came for a diagnosis. In some women, infertility may be the sole manifestation of CD. Had we been able to include all these women in our study, this is likely to have had a negative effect on the overall fertility rates in CD.
We based our CD diagnosis solely on biopsy findings, in which case positive CD serology was not required for the diagnosis. Still, a validation of a subset of patients with VA in this study found that 88% with available serology data had a positive CD serology at the time of biopsy.26 In clinical practice a diagnosis of CD is also based on the patient's symptoms and signs. We did not have complete data on symptoms, but through patient chart reviews we found that 79% of the patients with VA had gastrointestinal symptoms and that 95% of the patients with VA had CD.26
Finally, we used number of births rather than number of pregnancies as our fertility measure. Given that stillbirths and miscarriages are more common in CD,18 fertility would have been even higher in CD had we used pregnancy as our outcome measure.
In conclusion, women with CD had normal fertility, but their fertility was decreased in the last 2 years preceding diagnosis.
Funding JFL was supported by a grant from the Örebro University Hospital while writing this article. This project was supported by a grant from the Swedish Society of Medicine, the Swedish Research Council, the Sven Jerring Foundation, the Örebro Society of Medicine, the Karolinska Institutet, the Clas Groschinsky Foundation, the Juhlin Foundation, the Majblomman Foundation, Uppsala-Örebro Regional Research Council and the Swedish Coeliac Society. OA is supported by a grant from Karolinska Institutet/Stockholm County Council. OS is supported by a post-doctoral scholarship from the Swedish Society of Medicine.
Competing interests None declared.
Ethics approval This study was conducted with the approval of the Karolinska Institutet/Stockholm, Sweden.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.