Pregnant women were stratified according to their age in years at the index visit. For each group (women with qualifying pregnancies ranged in age from 15 to 38 years), a comparison cohort was constructed from among the women who had never been pregnant. To select a comparison group that was as similar as possible to the pregnant women, we defined an index visit for each comparison case as the last visit (within 180 days) before she reached her half-birthday. In addition to the index visit, the comparison cases were required to have at least one visit each in the 12-month during-pregnancy period and the 18-month follow-up period. This requirement ensured that there was a minimum of two visits in the during-pregnancy period for women in both the pregnant and the nonpregnant groups. read more
Because a comparison cohort was constructed separately for each annual age group, and data were available for individual women over a period of several years, it was possible for a single patient to contribute to the comparison cohort for more than one age group. Thus, a total of 13,259 comparison cases were included, representing 3,896 individual never-pregnant patients. The number of comparison cases including patients at each age varied from 146 to 1,556, and the ratio of comparison cases to pregnant patients was at least 19.7:1 (median, 57:1).
Outcome variables were summarized for each patient according to the predetermined time period (ie, baseline, during, or follow-up). Lung function measures (FEV1 percent predicted and FVC percent predicted) were presented as the average value for each time period. In the ESCF, each visit identifies whether the patient is sick or stable; for the purposes of lung function analyses, only the lung function values from visits during which the patient was stable were used. The number of exacerbations, hospitalizations, and clinic visits, as well as the frequency of therapies, including inhaled antibiotics, IV antibiotics, oral quin-olone antibiotics, and dornase alfa (Pulmozyme; Genentech Inc; South San Francisco, CA), were recorded for each patient during each time period. Similarly, the average weight in kilograms and the percentage of ideal body weight were calculated for each patient.
Analyses of variance were used to compare treatments and outcomes for the pregnant women with those of the never-pregnant women for each single year of age, treating the three periods as repeated measures within each patient. All analyses were performed using a statistical software package (SAS, version 8.2; SAS Institute; Cary, NC). All of the p values that were reported are two-tailed; values < 0.05 were considered to be statistically significant.