Women participating in Cohort Norway (CONOR) health surveys conducted between 1994 and 2003 [
19] were linked to the Medical Birth Registry of Norway (MBRN) [
20] for identification of all subsequent deliveries through to the end of 2012 using a unique identifier replaced by a project-specific pseudo-id for the women: details of which are presented elsewhere [
21]. Participants were also linked to pregnancies prior to CONOR participation for ascertainment of a prior preterm delivery among parous women. Due to age differences in recruitment to the health surveys, data for the current analyses came predominately from three CONOR surveys which included women of reproductive age: The HUBRO Study in Oslo (24%), The Tromsø Study (21%), and the Nord-Trøndelag Health Study (HUNT, 50%), and the remaining from diverse survey regions. The HUBRO Study in Oslo oversampled immigrants: 38% of the HUBRO Study participants were not born in Norway, otherwise the study sample primarily represents ethnic Norwegians.
We utilized CONOR questionnaires for descriptive baseline self-reported background information of mothers. The a-priori attributes of primary interest included self-reported frequency of past-year weekly light and vigorous LTPA and family history of stroke or of a myocardial infarction younger than 60 years of age in siblings and/or parents (first degree relatives). LTPA questions had four answer options (none, < 1 h., 1–2 h., and 3 or more hrs. per week). Vigorous activity was defined as any activity resulting in sweating and/or shortness of breath, whereas light activity was defined by the absence of sweating and/or shortness of breath. Vigorous LTPA was combined into three categories (< 1 h, 1–2 h and ≥ 3 h per week), whereas light LTPA was combined into two groups (< 3 h and ≥ 3 h per week) given that 1–2 h of light LTPA per week is below recommended activity levels.
CONOR also assessed baseline self-reported characteristics including health status (diabetes mellitus, chronic hypertension, and asthma), educational level (primary, secondary, or any college/university), daily cigarette smoking (yes vs. no), and marital status (married/cohabitation vs. other). Measurements including weight and height for calculating body mass index (BMI, kg/m2) and resting systolic and diastolic blood pressure were taken by trained personnel as part of the baseline CONOR assessments.
Medical birth registry of Norway (MBRN)
The MBRN is based on compulsory notification of all live- and stillbirths from week 16 of gestation through standardized forms completed by obstetric nurse midwife/birth attendants since 1967 [
20]. The MBRN provided data regarding gestational age, birth weight, and numerous medical risk factors for or associates of preterm delivery: pre-pregnancy thyroid condition, chronic hypertension, early and mid-pregnancy bleeding, placental abruption or previa, preeclampsia/eclampsia (PE), gestational hypertension, pregestational and gestational diabetes, congenital malformations and stillbirth deliveries, and whether pregnancy was a result of assisted reproductive technology (ART). Identification of preterm deliveries (early preterm < 34 weeks of gestation; late preterm ≥34 < 37 weeks of gestation) were based upon gestational age predicted by ultrasound when available. In this study, 17% of deliveries had missing data on gestational age by ultrasound. Thus, for these deliveries, we used date of last menstrual period to estimate gestational age.
Statistical analyses
Cohort descriptive characteristics are reported as percent or mean (SD). Multivariable logistic regression evaluated odds ratios (OR) and 95% confidence intervals (CI) for preterm delivery (< 37 weeks gestation), while multinomial logistic regression evaluated odds ratios (OR) for early preterm (< 34 weeks) and late preterm (34 through to end of 36 completed weeks gestation) relative to term deliveries. Births were the unit of analyses where mothers’ pseudo-id was entered as a cluster variable in regression analyses. There were four independent variables of interest: family history of stroke, family history of myocardial infarction prior to age 60 years, light LTPA and vigorous LTPA. Multivariable models considered important covariates. Model 1 included baseline age, length of time between CONOR participation and delivery, and parity risk category (nulliparous, parous with a prior preterm delivery, and the referent group of parous women with no prior preterm delivery). Model 2 included Model 1 covariates and baseline education (primary, secondary, or any college/university), daily cigarette smoking (yes vs. no), height (cm), and BMI status (underweight < 18.5, obese > 30, vs. referent group BMI of 18.5–29.9 kg/m2). Overweight (25–29.9 kg/m2) and normal weight (18.5–24.9 kg/m2) women were combined into one referent group as initial analyses showed no excess risk of premature deliveries to overweight relative to normal weight women. In the evaluation of LTPA, model 3 included model 2 covariates and both light and vigorous LTPA.
Supplementary analyses considered Model 3 covariates with medical conditions considered risk factors for or associates of preterm delivery: recipient of ART, a pre-pregnancy thyroid condition, one or more of the following three conditions combined (early or mid-pregnancy bleeding, placental abruption, or placenta previa), any congenital malformation, chronic hypertension and diabetes mellitus noted in CONOR health surveys and/or the MBRN, PE, or stillbirth deliveries. Further, supplemental analyses considered family history of stroke and myocardial infarction together with LTPA with model 3 covariates. Asthma reported in the baseline health surveys was not related to early, late or total preterm deliveries and was therefore not considered in these supplementary analyses.
A sensitivity analysis was conducted removing those with stillbirth delivery, chronic hypertension, PE, or diabetes (n = 920) as these conditions were considered potential intermediaries between physical activity and preterm delivery and between family histories and preterm deliveries. Also, as parous and nulliparous women may have different physical activity patterns, we conducted an additional sensitivity analyses of physical activity limited to nulliparous women at baseline. Finally, we conducted a sensitivity analysis removing those with a family history of stroke or of myocardial infarction.
While the Oslo region included immigrant women, we lacked data on immigrant status of participants for our analyses. Survey regions did not relate to risk of preterm or to LTPA patterns and was therefore not considered as a covariate in the analyses. However, while the mean and median heights were similar between the three survey regions, the Oslo region had a wider SD in heights likely reflecting the greater variability in women’s heights associated with a population which includes immigrant women from countries with shorter heights than that of Western Europe and Norway [
22]. The inclusion of height in our multivariable models was a means of adjusting, albeit indirectly, for immigrant status in addition to adjusting for the strong inverse association overall between maternal height and preterm delivery.
Data on light LTPA was missing for 6.0% (n = 788), and vigorous LTPA for 6.4% (n = 840) of participants. Otherwise, the percent missing data was low for most covariates: BMI, height, education (< 1.5%), and smoking (4.7%). Analyses were based upon complete data available for the independent variables and covariates. Statistical analysis was performed using Stata version 15 (StataCorp LP, 2017, College Station, TX, USA).