Databases

This study used 2 large-scale, nationwide, population-based datasets. The first dataset was sourced from the 2000-2003 National Health Insurance Research Dataset (NHIRD), published by the National Health Research Institute in Taiwan. The NHIRD consists of registries of contracted medical facilities and board-certified physicians along with inpatient and ambulatory care claims for more than 22 million enrollees, more than 98% of the island’s population. The NHIRD provides 1 principal diagnosis from the International Classification of Disease, Ninth Revision , Clinical Modification (ICD-9-CM) code and up to 4 secondary ICD-9-CM diagnoses for each patient.

The second database used in this study is obtained from the birth certificate registry published by Taiwan’s Ministry of the Interior. The data on birth certificates include both infants’ and parents’ birth dates, gestational week at birth, birthweight, sex, parity, place of birth, parental educational levels, and maternal marital status. The registration of all births is mandatory in Taiwan, and the completeness and validity of Taiwan’s birth registry has been verified.

With assistance from the Bureau of the National Health Insurance (NHI) in Taiwan, the mother’s and infant’s unique personal identification numbers provided links between the NHIRD and birth certificate data. Confidentiality assurances were addressed by abiding by the data regulations of the NHI. All personal identifiers were encrypted by the NHI before release to the researchers. Because the NHIRD consists of deidentified secondary data released to the public for research purposes, this study was exempt from full review by the internal review board.

Study sample

We identified a total of 473,529 women who had singleton births in Taiwan between Jan. 1, 2001, and Dec. 31, 2003. If a mother had more than 1 singleton birth during the study period, we selected only the first for the study sample. Of these women, 23,822 were identified as having visited ambulatory care clinics or outpatient departments of hospitals for treatment of peptic ulcer (ICD-9-CM code 531-533) during pregnancy.

Because many researchers question the coding validity of administrative databases, for the study cohort, we selected patients who had at least 3 consensus peptic ulcer diagnoses during pregnancy and who had undergone EGD test to confirm their PUD diagnosis within the 2 years preceding the index pregnancy. This left a total of 2120 women with PUD for analysis.

Our comparison cohort was extracted from the remaining 449,707 mothers. We randomly selected 10,600 women (5 for every mother with peptic ulcers) matched with the study group in terms of age (<20, 20-24, 25-29, 30-34, and ≥35 years) and year of delivery.

Variables of interest

The dependent variables were all dichotomous: whether an infant had LBW, preterm gestation, or was SGA. According the World Health Organization, the standard cutoff point for LBW is 2500 g (<2500 g, ≥2500 g). Preterm birth was defined as birth occurring at a gestational age less than 37 weeks, and SGA is defined as birthweight below the 10th percentile for gestational age. Savitz et al proposed a lack of concordance among these adverse fetal outcomes. Manifold outcome measures should be assessed, with the results from each measure examined separately. Our study thus adopted multiple outcomes for evaluation.

The key independent variable of interest was whether a mother had visited ambulatory care centers for the treatment of PUD during their pregnancy. A further dichotomous variable was generated for the women with PUD to distinguish those who received medications such as H ₂ blockers or proton pump inhibitors for more than 1 month during pregnancy and those who did not.

Other potential confounders contributing to pregnancy outcomes were also taken into consideration. These included characteristics of the infant (sex), mother (age, parity, the highest education level, and marital status), father (age and the highest education level), and family monthly income (including mothers’ and fathers’ monthly income). Parental age difference was also included because of its documented effects on birth outcomes, irrespective of any uniquely maternal characteristics.

Statistical analysis

The SAS statistical package (SAS System for Windows, version 8.2; SAS Institute, Inc., Chicago, IL) was used to perform all analyses in this study. Pearson χ ² tests were used to examine the differences in characteristics of mother, father, and infant comparing women with PUD and unaffected women. Multivariate logistic regression analyses were used to calculate the risk of LBW, preterm gestation, and SGA for these 2 cohorts. A significance level of .05 was selected to determine the significance of predictors in the models.