Abstract

BACKGROUND

Adverse pregnancy outcomes have been associated with maternal celiac disease (CD). In this study, we investigate the effect of treated and untreated maternal CD on infant birthweight and preterm birth.

METHODS

A population-based cohort study consisted of all singleton live births in Denmark between 1 January 1979 and 31 December 2004 was used. A total of 1 504 342 babies were born to 836 241 mothers during the study period. Of those, 1105 babies were born to women with diagnosed CD and 346 were born to women with undiagnosed CD. Women with diagnosed CD were considered as treated with a gluten free diet while women with undiagnosed CD were considered as untreated. The outcome measures were: birthweight, small for gestational age (SGA: birthweight <10th centile), very small for gestational age (VSGA: birthweight <5th centile) and preterm birth. We compared these measures in treated and untreated women with those of a reference group (no history of CD).

RESULTS

Women with untreated CD delivered smaller babies [difference = −98 g (95% CI: −130, −67)], with a higher risk of SGA infants [OR = 1.31 (95% CI: 1.06, 1.63)], VSGA infants [OR = 1.54 (95% CI: 1.17, 2.03)] and preterm birth [OR = 1.33 (95% CI: 1.02, 1.72)] compared with women without CD. Women with treated CD had no increased risk of reduced mean birthweight, risk of delivering SGA and VSGA infants or preterm birth compared with women without CD.

CONCLUSION

Untreated maternal CD increases the risk of reduced birthweight, the risk of delivering SGA and VSGA infants and preterm birth. Diagnosis and presumed treatment of maternal CD with a gluten-free diet appeared to result in a birthweight and preterm birth rate similar to those in women without CD.

Introduction

Celiac disease (CD) is a genetically determined, chronic, inflammatory intestinal disease induced by the ingestion of gluten (Green and Jabri, 2003; Green and Cellier, 2007). CD can be diagnosed at any age, but is most commonly identified in either early childhood or in the third or fourth decades of life (Van Heel and West, 2006). CD can exist in a very mild form and may go largely undetected as symptoms are often mild and vague such as fatigue often secondary to iron deficiency anemia. The classic signs of CD include diarrhoea with bulky foul smelling, floating stools due to steatorrhoea and flatulence. Malabsorption may occur resulting in weight loss, severe anemia and neurological disorders.

The true prevalence of CD is unknown but is estimated to affect around 1% of the population and is generally under-diagnosed (Rewers, 2005; Van Heel and West, 2006, Green and Cellier, 2007). Patients who are diagnosed with CD are instructed to follow a gluten-free diet which generally eases symptoms within days or weeks, although histological recovery can take months or even years (Green and Cellier, 2007).

Low birthweight and preterm birth are strong predictors of newborn mortality and morbidity which reaches far beyond the newborn period (McIntire et al., 1999; Lopez, 2001; Slattery and Morrison, 2002; Thornton et al., 2004; Gardosi et al., 2005). Adverse pregnancy outcomes such as reduced birthweight, low birthweight, small for gestational age (SGA), miscarriages and preterm birth have been associated with maternal CD (Ciacci et al., 1996; Norgard et al., 1999; Ludvigsson and Ludvigsson, 2001, Ludvigsson et al., 2005, Salvatore et al., 2007). In a population-based Swedish study, Ludvigsson et al. (2005) reported an association between undiagnosed maternal CD and low birthweight, SGA and preterm birth. Women with diagnosed CD, who were presumably treated before giving birth, had no increased risk of poor fetal growth compared with non-celiac women. This is in contrast, however, to an earlier Swedish study that reported both untreated and treated maternal CD increased the risk of low birthweight (Ludvigsson and Ludvigsson, 2001).

In the present study, we aimed to assess the effect of treated and untreated CD during pregnancy on birthweight, SGA and preterm birth using data from the entire Danish pregnant population over 25 years.

Materials and Methods

Using the Danish Medical Birth Register we identified all singleton live births in Denmark between 1 January1979 and 31 December 2004 and obtained information on their birthweight and gestational age (Knudsen and Olsen, 1998). By means of the Civil Registration Number, the children were linked to their mothers and legal fathers. The Civil Registration Number is a unique identifier used uniformly across services in Denmark and enables linkage between all the Danish national registers (Pedersen et al., 2006). This linkage enabled us to obtain information on parity, maternal and paternal age, infant sex and year of birth from the Civil Registration System. Information about CD in the mothers and fathers, and diabetes, hypertension, myocardial infarction and renal disease in the mothers was obtained from the Danish National Hospital Register (Anderson et al., 1999), which contains information on all inpatient discharges from Danish hospitals since 1977; outpatient attendances have been included in the register since 1995. All treatment is free of charge for Danish residents. Diagnostic information in the Danish National Hospital Register is based on the Danish version of the International Classification of Diseases, 8th revision (ICD-8) from 1977 to 1993, and the International Classification of Diseases, 10th revision (ICD-10) from 1994 to 2007. Mothers were classified as having CD if they had been hospitalized or in outpatient care with a diagnosis of CD (ICD-8: 269.00; ICD-10: K90.0). The time of onset was defined as the first day of the first contact to the hospital with the relevant diagnosis. Children were defined as exposed to a history of maternal diabetes, hypertension, myocardial infarction or renal disease in the mother if the mother was diagnosed with these conditions prior to the birth of the child.

Women were categorized in three groups according to the timing of the diagnosis of CD (women may appear more than once in the cohort if they had more than one pregnancy during 1979–2004). (i) Untreated women included those who were diagnosed after the date of birth of their child; we assume that untreated CD was present during any pregnancy which preceded the diagnosis. (ii) Treated women were those who were diagnosed with CD at least 90 days before the date of the beginning of gestation (date of the beginning of gestation was estimated as date of delivery minus gestational age). (iii) Women without CD included women who were never diagnosed with CD during 1977–2007; a small percentage of this group may have had undiagnosed disease but they were all assumed not to have CD. We assumed that any person who was diagnosed with CD adopted a gluten free diet and that the treatment took ∼90 days to take effect. Therefore, we excluded those who were diagnosed during the period from 90 days prior to pregnancy to delivery. We also excluded offspring of celiac fathers because treated paternal CD has been previously linked to lower birthweight and shorter gestational age (Ludvigsson and Ludvigsson, 2001). There were no children in the study cohort with two celiac parents.

We considered birthweight records entered in the register as over 5500 g or <500 g as invalid. Records of birthweight and gestational age were assumed invalid if birthweight was recorded as more than 1500 g and gestational age <29 weeks. For gestational age 29–33 weeks records were assumed invalid if birthweight was recorded as over 2800 g. Birthweight had been recorded in 10 g intervals between 1979 and 1990 and in 1 g intervals from 1991 onwards. However, due to a marked digit preference to the nearest 10, 50 and 100 g in the birthweight distribution we assumed that birthweight was recorded in 100 g during the study period.

Gestational age was based on the last menstrual period. This was often corrected by ultrasound measurements particularly in recent years (Pedersen et al., 2007). Gestational age was considered invalid if it was recorded as less than 23 or more than 44 weeks. Any baby with an invalid or missing birthweight or gestational age was excluded from the analysis.

We defined SGA as birthweight below the 10th percentile of the gestational age and sex-specific distributions. Due to observed increases in mean birthweight over the years, percentiles were also defined separately for four time periods: 1979–1986, 1987–1992, 1993–2000 and 2001–2004 for SGA determination. Very small for gestational age (VSGA) was defined as birthweight below the fifth percentile of the gestational age, sex-specific and time period distributions. Because birthweight was rounded to the nearest 100 g, cut-offs which included exactly 10 and 5% of weights could not be defined; the nearest lower figures were used instead giving percentages closer to 8 and 4%. Preterm birth was defined as delivery before 37 gestation weeks.

Study cohort

During the study period there were records of 1 563 322 children in the Danish Medical Birth Register. Of these, 29 children were excluded because their mothers were diagnosed during or up to 90 days before pregnancy and 620 because they had celiac fathers. A further 58 331 children were excluded because either their birthweight or gestational age were missing or invalid. The final cohort consisted of 1 504 342 children. We decided to exclude all births between 1 January 1973 and 31 December 1978 from the study cohort because birthweight was recorded in 250 g intervals during that period. Moreover between 1973 and 1977 gestational age of term babies (≥37 weeks) was recorded in one category.

Statistical analysis

We used linear regression analysis (Montgomery et al., 2001) to examine the association between treated and untreated maternal CD and birthweight. The models were adjusted for parity, infant sex, maternal age, maternal medical history of diabetes, hypertension, myocardial infarction and renal disease, infant year of birth and paternal age, all of which were represented as categorical variables (Table I). We also adjusted for gestational age, as a continuous variable. Since the relationship between gestational age and birthweight is not linear we used fractional polynomial regression to determine the best functional relationship resulting in a model with four polynomial functions of gestational age. Because birthweight variance was found to vary by gestational age week and the lack of independence in the data we used linear regression analysis with ‘robust’ variance estimator (White, 1980).

Table I

Maternal characteristics in the Danish Medical Birth Registry 1979–2004 in relation to CD (singleton births only)

Variable Non-celiac (n = 1502891), n (%) Untreated CD (n = 1105), n (%) Treated CD (n = 346), n (%) 
Maternal age 
 <20 40 147 (2.7) 36 (3.3) 4 (1.6) 
 20–24 30 4170 (20.2) 246 (22.3) 42 (12.1) 
 25–29 582 993 (38.8) 441 (39.9) 127 (36.7) 
 30–34 413 609 (27.5) 289 (26.1) 121 (35.0) 
 35–39 141 059 (9.4) 78 (7.1) 41 (11.8) 
  ≥ 40 20 913 (1.4) 15 (1.4) 11 (3.2) 
 Mean (SD) 28.2 (4.9) 27.8 (4.8) 29.7 (4.8) 
Paternal age 
  < 25 164 003 (10.9) 143 (12.9) 26 (7.5) 
 25–29 472 595 (31.4) 382 (34.6) 100 (28.9) 
 30–34 493 934 (32.9) 352 (32.8) 122 (35.3) 
 35–39 248 335 (16.5) 163 (14.7) 62 (17.9) 
  ≥ 40 124 024 (8.3) 65 (5.9) 36 (10.4) 
 Mean (SD) 31.0 (5.8) 30.4 (5.5) 32.0 (6.1) 
Parity 
 1 673 451 (44.8) 491 (44.4) 150 (43.3) 
 2 561 493 (37.4) 412 (37.3) 120 (34.7) 
  ≥ 3 267 947 (17.8) 202 (18.3) 76 (22.0) 
Infant sex 
 Male 771 655 (51.3) 549 (49.7) 178 (51.4) 
 Female 731 236 (48.7) 556 (50.3) 168 (48.5) 
Maternal hypertension, myocardial infarction or renal disease 
 Yes 13 487 (0.9) 9 (0.8) 11 (3.2) 
 No 1 489 404 (99.1) 1096 (99.2) 335 (96.8) 
Maternal diabetes 
 Yes 17 589 (1.2) 26 (2.3) 20 (5.8) 
 No 1 485 302 (98.8) 1079 (97.7) 326 (94.2) 
Smoking (1997–2004) 
 Non-smoker 372 339 (75.5) 194 (78.9) 181 (77.7) 
 Smoker 102 866 (20.8) 39 (15.8) 44 (18.9) 
 Unknown 18 123 (3.7) 13 (5.3) 8 (3.4) 
Variable Non-celiac (n = 1502891), n (%) Untreated CD (n = 1105), n (%) Treated CD (n = 346), n (%) 
Maternal age 
 <20 40 147 (2.7) 36 (3.3) 4 (1.6) 
 20–24 30 4170 (20.2) 246 (22.3) 42 (12.1) 
 25–29 582 993 (38.8) 441 (39.9) 127 (36.7) 
 30–34 413 609 (27.5) 289 (26.1) 121 (35.0) 
 35–39 141 059 (9.4) 78 (7.1) 41 (11.8) 
  ≥ 40 20 913 (1.4) 15 (1.4) 11 (3.2) 
 Mean (SD) 28.2 (4.9) 27.8 (4.8) 29.7 (4.8) 
Paternal age 
  < 25 164 003 (10.9) 143 (12.9) 26 (7.5) 
 25–29 472 595 (31.4) 382 (34.6) 100 (28.9) 
 30–34 493 934 (32.9) 352 (32.8) 122 (35.3) 
 35–39 248 335 (16.5) 163 (14.7) 62 (17.9) 
  ≥ 40 124 024 (8.3) 65 (5.9) 36 (10.4) 
 Mean (SD) 31.0 (5.8) 30.4 (5.5) 32.0 (6.1) 
Parity 
 1 673 451 (44.8) 491 (44.4) 150 (43.3) 
 2 561 493 (37.4) 412 (37.3) 120 (34.7) 
  ≥ 3 267 947 (17.8) 202 (18.3) 76 (22.0) 
Infant sex 
 Male 771 655 (51.3) 549 (49.7) 178 (51.4) 
 Female 731 236 (48.7) 556 (50.3) 168 (48.5) 
Maternal hypertension, myocardial infarction or renal disease 
 Yes 13 487 (0.9) 9 (0.8) 11 (3.2) 
 No 1 489 404 (99.1) 1096 (99.2) 335 (96.8) 
Maternal diabetes 
 Yes 17 589 (1.2) 26 (2.3) 20 (5.8) 
 No 1 485 302 (98.8) 1079 (97.7) 326 (94.2) 
Smoking (1997–2004) 
 Non-smoker 372 339 (75.5) 194 (78.9) 181 (77.7) 
 Smoker 102 866 (20.8) 39 (15.8) 44 (18.9) 
 Unknown 18 123 (3.7) 13 (5.3) 8 (3.4) 

CD = celiac disease; SD = standard deviation.

Note: There were 835 456 non-celiac mothers, 559 untreated celiac mothers and 226 treated mothers.

The associations between maternal CD and SGA, VSGA and preterm birth, were examined using multiple logistic regression. The models for SGA, VSGA and preterm birth were adjusted for parity, maternal age, maternal medical history of diabetes, hypertension, myocardial infarction and renal disease, infant year of birth and paternal age. Because of the lack of independence in the data we used ‘robust’ variance estimator.

The data analyses estimated the differences in pregnancy outcomes between (i) offspring of mothers with untreated CD and mothers without CD and between (ii) offspring of mothers with treated CD and mothers without CD using a single model with a three category CD variable: no-CD, untreated and treated, with no-CD used as the reference category. Additional analyses on birthweight and SGA were performed by restricting the data to term babies. We also examined whether the effect of CD on birthweight and SGA was the same for nulliparous and multiparous women. This was done by adding a term representing the statistical interaction between CD and parity (nulliparous or multiparous).

To examine the potential confounding effect of maternal smoking on the association between CD and birthweight we restricted the cohort to singleton live births between 1 January 1997 and 31 December 2004 when smoking information was recorded in the Danish National Hospital Register. We estimated the effect of CD on birthweight adjusted for maternal smoking, maternal and paternal age, parity, infant sex, year of birth, gestational age, maternal medical history of diabetes and maternal medical history of hypertension, myocardial infarction and renal disease. We then repeated the same analysis omitting maternal smoking. All statistical analyses were performed using Stata Software 10 (Stata Reference Manual: Release 10. College Station, 2007).

The study was approved by the Danish Data Protection Agency and the Danish National Board of Health.

Results

The study cohort consisted of 1 504 342 deliveries (836 241 women). During the study period 785 (0.094%) mothers were diagnosed with CD and they had 1451 babies. A total of 346 babies were born following their mothers’ diagnosis and were presumed to have been treated for CD while 1105 babies were born before their mothers’ diagnosis and treatment of CD. Mean birthweight was 3503 g in the offspring of treated mothers, 3354 g in the offspring of untreated mothers and 3490 g in the offspring of mothers without CD. Women with untreated CD and their partners appeared to be slightly younger than couples in which the woman had treated CD or no history of CD. Women with treated CD were more likely to have diabetes, hypertension, myocardial infarction and renal disease (Table I).

Birthweight in women with untreated CD versus women without CD

The estimates of the association between untreated CD and birthweight are presented in Table II, column 3. The adjusted analyses suggest that untreated CD reduced mean birthweight by 100 g [adjusted difference = −98 (95% CI: −130, −67)]. Similar results were obtained after restricting the analysis to term babies [adjusted difference = −100 (95% CI: −132, −68)]. Untreated CD appeared to reduce mean birthweight by 70 g in nulliparas women [adjusted difference = −70 (95% CI: −108, −31)] and 121 g in multiparous women [adjusted difference = −121 (95% CI: −160, −82)].

Table II

The effect of untreated and treated maternal CD on birthweight

Variable Non-celiac women Women with untreated CD Women with treated CD 
N births (%)c 1 502 891 (99.91) 1105 (0.07) 346 (0.02) 
Birthweight 
 Mean birthweight in grams (SD) 3490 (569) 3354 (579) 3503 (546) 
 Adjusted birthweight differencea, g (95% CI) Reference −98 (−130, −67) −15 (−70, 41) 
 N term births 1 433 516 1038 329 
 Adjusted birthweight differencea, g (95% CI) in term babies Reference −100 (−132, −68) −29 (−83, 25) 
Nulliparous women 
N births of nulliparous mothers (%)c 673 451 (99.9) 491 (0.07) 150 (0.02) 
 Adjusted birthweight differenceb, g (95% CI) in offspring of nulliparous mothers Reference −70 (−108, −31) −21 (−91, 50) 
Multiparous women 
N births of multiparous mothers (%)c 829 440 (99.9) 614 (0.07) 196 (0.02) 
 Adjusted birthweight differenceb, g (95% CI) in offspring of multiparous mothers Reference −121(−160, −82) −10 (−81, 62) 
Variable Non-celiac women Women with untreated CD Women with treated CD 
N births (%)c 1 502 891 (99.91) 1105 (0.07) 346 (0.02) 
Birthweight 
 Mean birthweight in grams (SD) 3490 (569) 3354 (579) 3503 (546) 
 Adjusted birthweight differencea, g (95% CI) Reference −98 (−130, −67) −15 (−70, 41) 
 N term births 1 433 516 1038 329 
 Adjusted birthweight differencea, g (95% CI) in term babies Reference −100 (−132, −68) −29 (−83, 25) 
Nulliparous women 
N births of nulliparous mothers (%)c 673 451 (99.9) 491 (0.07) 150 (0.02) 
 Adjusted birthweight differenceb, g (95% CI) in offspring of nulliparous mothers Reference −70 (−108, −31) −21 (−91, 50) 
Multiparous women 
N births of multiparous mothers (%)c 829 440 (99.9) 614 (0.07) 196 (0.02) 
 Adjusted birthweight differenceb, g (95% CI) in offspring of multiparous mothers Reference −121(−160, −82) −10 (−81, 62) 

aAdjusted for gestational age, parity, maternal age, paternal age, infant sex and year of birth, maternal medical history of diabetes, maternal medical history of hypertension, myocardial infarction and renal disease.

bAdjusted for gestational age, maternal age, paternal age, infant sex and year of birth, maternal medical history of diabetes, maternal medical history of hypertension, myocardial infarction and renal disease.

cPercentages are based on figures within the same row.

When we restricted the cohort to the 1997–2004 births, untreated CD appeared to reduce mean birthweight by 130 g compared with birthweight in women without CD [adjusted difference = −130 (95% CI: −190, −70)]. When we adjusted the same model for maternal smoking in addition to the other variables the estimate changed slightly [adjusted difference = −140 (95% CI: −200, −80)]. Furthermore, excluding offspring of women who had a medical history of diabetes, hypertension, acute myocardial infarction or renal disease from the data did not change the conclusions.

SGA, VSGA and preterm birth in women with untreated CD versus women without CD

The logistic regression models suggested an association between untreated CD and SGA [adjusted OR = 1.31 (95% CI: 1.06, 1.63)] (Table III). Restricting the analysis to term births did not change the result materially [adjusted OR = 1.27 (95% CI: 1.02, 1.58)]. Untreated CD was associated with SGA in multiparous [adjusted OR = 1.57 (95% CI: 1.19, 2.09)] but not nulliparous women [adjusted OR = 1.10 (95% CI: 0.83, 1.46)]. Moreover, untreated CD was associated with VSGA [adjusted OR = 1.54 (95% CI: 1.17, 2.03)] and preterm birth [adjusted OR = 1.33 (95% CI: 1.02, 1.72)].

Table III

The effect of untreated and treated CD on SGA, VSGA and preterm birth

Variable Non-celiac women Women with untreated CD Women with treated CD 
N births 1502891 1105 346 
SGA 
 Number of SGA births (%) 122677 (8.16) 115 (10.41) 28 (8.10) 
 Crude SGA OR (95% CI) Reference 1.31 (1.06, 1.62) 0.99 (0.66, 1.48) 
 Adjusted SGA OR (95% CI)a Reference 1.31 (1.06, 1.63) 1.01 (0.67, 1.50) 
 Number of SGA births in term babies (%)c 116 896 (8.15) 105 (10.12) 27 (8.21) 
 Adjusted SGA OR (95% CI) in term babiesa Reference 1.27 (1.02, 1.58) 1.06 (0.71, 1.58) 
Nulliparous women, n 673 451 491 150 
Number of SGA babies in nulliparous women (%)c 69 592 (10.33) 55 (11.20) 19 (12.7) 
Adjusted SGA OR (95% CI) in offspring of nulliparous mothersb Reference 1.10 (0.83, 1.46) 1.23 (0.76, 2.00) 
Multiparous women, n 829 440 614 196 
Number of SGA babies in multiparous women (%)c 53 085 (6.40) 60 (9.77) 9 (4.59) 
Adjusted SGA OR (95% CI) in offspring of multiparous mothersb Reference 1.58 (1.19, 2.09) 0.74 (0.38, 1.45) 
VSGA 
 Number of VSGA births (%) 62 324 (4.15) 69 (6.24) 14 (4.05) 
 Crude VSGA OR (95% CI) Reference 1.54 (1.17, 2.02) 0.97 (0.55, 1.72) 
 Adjusted VSGA OR (95% CI)a Reference 1.54 (1.17, 2.03) 1.00 (0.59, 1.78) 
 Number of VSGA births in term babies (%)c 59 541 (4.15) 65 (6.26) 14 (4.25) 
 Adjusted VSGA OR (95% CI) in term babiesa Reference 1.54 (1.17, 2.03) 1.05 (0.59, 1.86) 
Preterm birth 
 Number of preterm births (%) 69 407 (4.62) 67 (6.06) 17 (4.91) 
 Crude OR (95% CI) Reference 1.33 (1.03, 1.73) 1.07 (0.66, 1.72) 
 Adjusted OR(95% CI)a Reference 1.33 (1.02, 1.72) 0.92 (0.58, 1.46) 
Variable Non-celiac women Women with untreated CD Women with treated CD 
N births 1502891 1105 346 
SGA 
 Number of SGA births (%) 122677 (8.16) 115 (10.41) 28 (8.10) 
 Crude SGA OR (95% CI) Reference 1.31 (1.06, 1.62) 0.99 (0.66, 1.48) 
 Adjusted SGA OR (95% CI)a Reference 1.31 (1.06, 1.63) 1.01 (0.67, 1.50) 
 Number of SGA births in term babies (%)c 116 896 (8.15) 105 (10.12) 27 (8.21) 
 Adjusted SGA OR (95% CI) in term babiesa Reference 1.27 (1.02, 1.58) 1.06 (0.71, 1.58) 
Nulliparous women, n 673 451 491 150 
Number of SGA babies in nulliparous women (%)c 69 592 (10.33) 55 (11.20) 19 (12.7) 
Adjusted SGA OR (95% CI) in offspring of nulliparous mothersb Reference 1.10 (0.83, 1.46) 1.23 (0.76, 2.00) 
Multiparous women, n 829 440 614 196 
Number of SGA babies in multiparous women (%)c 53 085 (6.40) 60 (9.77) 9 (4.59) 
Adjusted SGA OR (95% CI) in offspring of multiparous mothersb Reference 1.58 (1.19, 2.09) 0.74 (0.38, 1.45) 
VSGA 
 Number of VSGA births (%) 62 324 (4.15) 69 (6.24) 14 (4.05) 
 Crude VSGA OR (95% CI) Reference 1.54 (1.17, 2.02) 0.97 (0.55, 1.72) 
 Adjusted VSGA OR (95% CI)a Reference 1.54 (1.17, 2.03) 1.00 (0.59, 1.78) 
 Number of VSGA births in term babies (%)c 59 541 (4.15) 65 (6.26) 14 (4.25) 
 Adjusted VSGA OR (95% CI) in term babiesa Reference 1.54 (1.17, 2.03) 1.05 (0.59, 1.86) 
Preterm birth 
 Number of preterm births (%) 69 407 (4.62) 67 (6.06) 17 (4.91) 
 Crude OR (95% CI) Reference 1.33 (1.03, 1.73) 1.07 (0.66, 1.72) 
 Adjusted OR(95% CI)a Reference 1.33 (1.02, 1.72) 0.92 (0.58, 1.46) 

aAdjusted for parity, maternal age, paternal age, infant year of birth, maternal medical history of diabetes, maternal medical history of hypertension, myocardial infarction and renal disease.

bAdjusted for maternal age, paternal age, infant year of birth, maternal medical history of diabetes, maternal medical history of hypertension, myocardial infarction and renal disease.

cThis percentage is based on the relevant populations i.e. term babies, offspring of nulliparous women and offspring of multiparous women.

CD = celiac disease; CI = confidence intervals; OR = odds ratio; SGA = small for gestational age; VSGA = very small for gestational age.

Women with treated CD versus women without CD

There was no evidence to suggest an association between treated CD and a reduction in birthweight [adjusted difference = −15 (95% CI: −70, 41)] (Table II). The results did not suggest an association between treated CD and SGA [adjusted OR = 1.01 (95% CI: 0.67, 1.50)], VSGA [adjusted OR = 1.00 (95% CI: 0.57, 1.78)] or preterm birth [adjusted OR = 0.92 (95% CI: 0.58, 1.46)] (Table III).

Discussion

This study suggests that untreated CD during pregnancy increased the risk of reduced mean birthweight, SGA, VSGA and preterm birth. In contrast, we found no association between treated CD during pregnancy and birthweight or preterm birth. Thus the treatment of women with CD seems to reverse the effects of CD on birthweight and preterm birth to a level similar to that in women without CD. The association between untreated CD and birthweight, SGA and VSGA existed even after excluding preterm births. The effect of untreated CD on birthweight appears to be greater in multiparous women compared with first time mothers.

The present study is population-based and the second largest to date on this topic. We were able to adjust for several potential confounders and stratify or restrict the data to rule out other potential confounding factors such as co-morbidity and smoking. By using data from the Danish national registers we avoided patient selection and recruitment biases and loss to follow-up. We had almost complete information on gestational age and birthweight. Kristensen et al. (1996) assessed the validity of gestational age data by comparing the information on gestational age in the Medical Birth Register and the medical records using 1982–1987 data. The authors found an 87% agreement between the two sources when gestational age was measured in weeks. However, records of gestational age may suffer from misclassification in part due to different methods of calculation and previous research suggested that preterm births are more likely to have been misclassified (Gjessing et al., 1999). It could be argued that women with treated CD had better antenatal care and were followed more closely and therefore their gestational age data were less likely to be misclassified compared with untreated celiac women. However, to our knowledge, women with treated CD in pregnancy do not receive different antenatal care compared with pregnant women without CD. Therefore misclassification of gestational age is unlikely to have occurred at a different rate among women with and without CD in our study and should not affect the presented estimates.

Although the present study is population-based and adjusted for several potential confounders there are some limitations. We had information on maternal smoking for those who were born between 1997 and 2004 only. Maternal cigarette smoking is associated with adverse pregnancy outcomes (Delpisheh et al., 2007) and previous research reported an inverse association between smoking and risk of CD (Snook et al., 1996). Our sensitivity analyses showed that adjustment for maternal smoking had little effect on the reported estimates.

We extracted the information on CD diagnoses from the Danish National Hospital Register. It is likely, given the subclinical nature of CD in many instances, that some women with CD may have been misclassified as non-celiac in the study cohort. The diagnosis of CD is recorded following discharge from, or at the time of outpatient visits at, the hospital and not by screening programs of CD. The Hospital register has been effective from 1977 therefore we will have missed diagnoses of CD before that date. It is widely accepted that clinical CD represents the tip of the iceberg and many patients may have symptoms for a long time and are admitted to hospital several times and undergo surgical procedures before they are diagnosed with CD (Green and Cellier, 2007). Only 0.09% of the women in the present study were diagnosed with CD (treated or untreated) although the true prevalence in other European populations is estimated to be close to 1% (Rewers, 2005; Green and Cellier, 2007). On the basis of a 1% true prevalence and 0.09% diagnosed cases of CD (785 women) it is possible that almost 8000 women with CD remain undiagnosed and in our study have been misclassified as non-celiac. However, since this is a tiny proportion of our reference group (women without CD), misclassification of women with CD as non-celiac is unlikely to greatly affect the reported risk estimates: it would tend to lead to underestimation rather than an overestimation of the effect of CD. In contrast, if women who have delivered smaller babies had a better chance of having their CD diagnosed than women who had a baby with an average birthweight, the effect of CD on birthweight could have been overestimated. However, it is not routine clinical practice to test for maternal CD following poor pregnancy outcomes. It is also possible that patients with other diseases may be registered as CD patients more often than patients without any other diseases but this is unlikely to greatly affect the reported risk estimates.

Our findings are consistent with the majority of the existing literature (Norgard et al., 1999, Ludvigsson et al., 2005), although the magnitude of the effect that we observed is lower than that reported in some of the previous studies (Ciacci et al., 1996; Salvatore et al., 2007). Our results are similar to those reported by Ludvigsson et al. (2005), which was based on the Swedish national registers. They reported a 60% increase in risk of SGA (defined as birthweight for gestational age below −2 standard deviations of the mean). Our results are also similar to those reported by Norgard et al. They used the Danish Medical Birth registry data between 1977 and 1992 and found that the mean birthweight of babies born to women with CD prior to their first hospitalization was 238 g [(95% CI) = 150, 325 g) lower than those of the control women. However, the results reported by Norgard et al., were based on data of 211 babies born to 127 celiac mothers although the present study was based on data of 1451 babies born to 785 celiac mothers.

We found that diagnosis of CD before conception and presumed treatment with gluten free diet improves the pregnancy outcome. This finding is consistent with the recent Swedish study (Ludvigsson et al., 2005) and others (Ciacci et al., 1996; Norgard et al., 1999) but contrasts with an earlier Swedish study which reported that even treated maternal CD was associated with lower birthweight (Ludvigsson and Ludvigsson, 2001). However, the early study (Ludvigsson and Ludvigsson, 2001) was based on only 53 live births of mothers with CD, although the later study (Ludvigsson et al., 2005) was based on 2078 offspring of celiac mothers.

The exact mechanisms whereby untreated maternal CD results in preterm delivery or smaller babies are unknown. It has been hypothesized that the adverse effect of untreated CD on birthweight and risk of preterm delivery may be due to nutritional deficiencies as a direct consequence of malabsorption. However, several studies have suggested that malabsorption or malnutrition is not a consistent feature in women with CD and adverse pregnancy outcome, nor does it appear that severity of symptoms is related to the degree of adverse outcome (Ciacci et al., 1996). Another possible mechanism may be the direct adverse effect of maternal auto antibodies, such as anti-tissue transglutaminase, on placental function (Hager et al., 1999). The presence of these antibodies at the feto-placental interface may result in compromised nutrient transfer, altered cell dynamics, suppression of growth factors and altered secretory function. We found the association between untreated CD and birthweight measures to be more pronounced in women who had previously given birth. In general, birthweight for gestational age increases from first to second pregnancy (Beaty et al., 1997). Whether our finding is due to a true interaction between parity and untreated CD needs further studies as this was not part of our prior hypotheses.

Untreated CD in pregnancy is associated with reduced infant birthweight and increased risks of SGA, VSGA and preterm birth. This study also supports the evidence that treatment of CD reduces the risk of reduced birthweight to that of the general population. Whether pre-pregnancy or antenatal screening of all women and the commencement of gluten free diets of those found to have CD would reduce these risks remains to be shown; however, it seems likely considering our results.

Funding

This study was funded by the Health Research Board, Ireland (Ref: CSA/2007/2).

Acknowledgements

We thank Professor Fergus Shanahan and Professor Eamonn Quigley, Department of Medicine, University College Cork for their comments and advice during the preparation of this manuscript.

References

Anderson
F
Madsen
M
Jorgensen
J
MellemkjÆR
L
Olsen
JH
The Danish National Hospital Register
Dan Med Bull
 , 
1999
, vol. 
46
 (pg. 
263
-
268
)
Beaty
T
Skjaerven
R
Breazeale
D
Liang
K
Analyzing sibship correlations in birthweight using large sibships from Norway
Genet Epidemiol
 , 
1997
, vol. 
14
 (pg. 
423
-
433
)
Ciacci
C
Cirillo
M
Auriemma
G
di Dato
G
Sabbatini
F
Mazzacca
G
Celiac disease and pregnancy outcome
Am J Gastroenterol
 , 
1996
, vol. 
91
 (pg. 
718
-
722
)
Delpisheh
A
Kelly
Y
Rizwan
S
Attia
E
Drammond
S
Brabin
B
Population attributable risk for adverse pregnancy outcomes related to smoking in adolescents and adults
Public Health
 , 
2007
, vol. 
121
 (pg. 
861
-
868
)
Gardosi
J
Kady
SM
Mcgeown
P
Francis
A
Tonks
A
Classification of stillbirth by relevant condition at death (ReCoDe): population based cohort study
Br Med J
 , 
2005
, vol. 
331
 (pg. 
1113
-
1117
)
Gjessing
HK
Skjærven
R
Wilcox
AJ
Errors in gestational age: evidence of bleeding early in pregnancy
Am J Public Health
 , 
1999
, vol. 
89
 (pg. 
213
-
218
)
Green
P
Cellier
C
Celiac disease
NEJM
 , 
2007
, vol. 
357
 (pg. 
1731
-
1743
)
Green
P
Jabri
B
Coeliac disease
Lancet
 , 
2003
, vol. 
362
 
Hager
H
Gliemann
J
Hamilton-dutoit
S
Ebbesen
P
Koppelhus
U
Jensen
P
Developmental regualtion of tissue transglutaminase during human placentation and expression in neoplastic trophoblast
J Pathol
 , 
1999
, vol. 
181
 (pg. 
106
-
110
)
Knudsen
LB
Olsen
J
The Danish medical birth registry
Dan Med Bull
 , 
1998
, vol. 
45
 (pg. 
320
-
323
)
Kristensen
J
Langhoff-ross
J
Skovgaard
LT
Kristensen
FB
Validation of the medical birth registration
J Clin Epidemiol
 , 
1996
, vol. 
49
 (pg. 
893
-
897
)
Lopez
BA
Overview of current research in parturition
Exp Physiol
 , 
2001
, vol. 
86
 (pg. 
213
-
222
)
Ludvigsson
JF
Ludvigsson
J
Coeliac disease in the father affects the newborn
Gut
 , 
2001
, vol. 
49
 (pg. 
169
-
175
)
Ludvigsson
J
Montgomery
S
Ekbom
A
Celiac disease and risk of adverse fetal outcome: a population-based cohort study
Gastroenterology
 , 
2005
, vol. 
129
 (pg. 
454
-
463
)
Mcintire
DD
Bloom
SL
Casey
BM
Leveno
KJ
Birth weight in relation to morbidity and mortality among newborn infants
N Engl J Med
 , 
1999
, vol. 
340
 (pg. 
1234
-
1238
)
Montgomery
DC
Peck
EA
Vining
GG
Introduction to Linear Regression Analysis
 , 
2001
New York
Wiley
Norgard
B
Fonager
K
Sorensen
HT
Olsen
J
Birth outcomes of women with celiac disease: a nationwide historical cohort study
Am J Gastroenterol
 , 
1999
, vol. 
94
 (pg. 
2435
-
2440
)
Pedersen
CB
Gøtzsche
H
Møller
Mortensen
PB
The Danish civil registration system
Dan Med Bull
 , 
2006
, vol. 
53
 (pg. 
441
-
449
)
Pedersen
CB
Sun
Y
Vestergaard
M
Olsen
J
Basso
O
Assessing fetal growth impairments based on family data as a tool for indentifying high-risk babies. An example with neonatal mortality
BMC Pregnancy Childbirth
 , 
2007
, vol. 
7
 pg. 
28
 
Rewers
M
Epidemiology of celiac disease: what are the prevalence, incidence and progression of celiac disease?
Gastroenterology
 , 
2005
, vol. 
128
 (pg. 
S47
-
S51
)
Salvatore
S
Finazzi
S
Radaelli
G
Lotzniker
M
Zuccotti
GV
Prevalence of undiagnosed celiac disease in the parents of preterm and/or small for gestational age infants
Am J Gastroenterol
 , 
2007
, vol. 
102
 (pg. 
168
-
173
)
Slattery
MM
Morrison
JJ
Preterm delivery
Lancet
 , 
2002
, vol. 
360
 (pg. 
1489
-
1497
)
Snook
J
Dwyer
L
Lee-elliott
C
Khan
S
Wheeler
D
Nicholas
D
Adult coeliac disease and cigarette smoking
Gut
 , 
1996
, vol. 
39
 (pg. 
60
-
62
)
Stata Reference Manual: Release 10
2007
College Station, TX
Stata Corp
Thornton
JG
Hornbuckle
J
Vail
A
Spiegelhalter
DJ
Levene
M
Infant wellbeing at 2 years of age in the Growth Restriction Intervention Trial (GRIT): multicentred randomised controlled trial
Lancet
 , 
2004
, vol. 
364
 (pg. 
513
-
580
)
Van heel
D
West
J
Recent advances in coeliac disease
Gut
 , 
2006
, vol. 
55
 (pg. 
1037
-
1046
)
White
H
A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskidasticity
Econometrica
 , 
1980
, vol. 
48
 (pg. 
817
-
838
)