Contribution of prepregnancy body mass index and gestational weight gain to caesarean birth in Canada

This study used data from the Public Health Agency of Canada’s Canadian Maternity Experiences Survey (MES). The MES was a cross-sectional survey of a stratified random sample of women who had a singleton live birth in Canada between November 2005 and May 2006. Women were identified using recent births drawn from a Census-based sampling frame. Women were eligible for the study if they were at least 15 years of age and were living with their infant at the time of data collection. Women living on First Nations reserves or in institutions were excluded. Data were collected by female interviewers between October 2006 and January 2007 using a computer-assisted telephone interview application. The majority (97%) of interviews were conducted between five and nine months postpartum. Out of 8,244 eligible women, 6,421 (78%) agreed to participate. In consideration of the sample design and non-participation, each MES record was assigned a sampling weight. The 6,421 respondents were thus weighted to represent 76,508 women nationally who had a singleton live birth between November 2005 and May 2006. Survey questions covered a broad range of pregnancy, birth and postpartum experiences. All data were based on women’s reports. Detailed information on the survey’s development, methodology and content has been reported elsewhere [17].

We excluded women with missing information on BMI or GWG (n = 79), as these were the principal determinants of interest. Mode of birth information was complete. We also excluded preterm births (< 37 weeks gestation) (n = 568) and women who were less than 18 years old (n = 183), as the BMI classification used was derived for ages 18 and older. These exclusions resulted in a final sample of 5,591 women weighted to represent 67,058 women.

The primary outcome was caesarean birth (planned and unplanned). Caesarean births were classified as planned if the decision about the mode of birth was made before the woman went into labour.

Prepregnancy BMI and GWG were the principal exposures of interest. They were derived from the following questions:

We categorized women according to the World Health Organization (WHO) standard as either being underweight (BMI < 18.5 kg/m²), normal weight (18.5 ≤ BMI < 25), overweight (25 ≤ BMI < 30) or obese (BMI ≥ 30). Women were also classified according to the Institute of Medicine’s recommended GWG ranges (Table 1) [18], as having gained above, within or below the recommended weight for their BMI.

We studied additional reproductive, health care, sociodemographic and psychosocial characteristics as potential confounders of the association between BMI, GWG and mode of birth. Birthweight-for-gestational-age was derived using a Canadian reference to categorize infants below the standard 10^th percentile as small-for-gestational age (SGA) and those above the standard 90^th percentile as large-for-gestational age (LGA) [19]. Sociodemographic variables assessed included the household’s low income cut-off level (LICO), which is a measure of the income threshold below which a family will likely spend 20 percentage points more than the average family on food, shelter and clothing [20]. Ethnicity was based on mother’s country of birth, grouped according to world regions; mothers born in Canada were categorized as Aboriginal off-reserve and non-Aboriginal. The MES questions for variables whose definitions are not self-evident are indicated in Table 2. Categorizations (for non-dichotomized variables) used in analyses are indicated in Table 3 in the results section.

Percentages were used to report observed distributions of BMI and GWG across maternal characteristics. We calculated adjusted odds ratios (ORs) for having a caesarean birth using multivariable logistic regression. With the exception of maternal age, all variables were treated as categorical in regression models. Records with missing values for covariates other than LICO were excluded from models (< 4%). Due to a larger number of missing LICO values (8.0%), a missing category was included for this variable. We calculated ORs across BMI and GWG groups for caesarean births overall as well as for unplanned and planned caesarean births. Normal BMI and within recommended GWG were the reference groups.

BMI and GWG were included in all multivariable models in order to estimate their independent associations (ORs) with caesarean birth. Other covariates were selected into models purposefully using the following steps [21]. Based on the Wald test from univariable logistic regression models, we initially included any variable with a p-value below 0.25. Covariates were then removed from the model if they were statistically non-significant and not a confounder. Significance was evaluated at the 0.05 level and confounding as a change of 15% or higher in the effect of BMI or GWG on the mode of birth outcome being modeled. To address significant interaction between parity, prior caesarean, BMI, GWG and mode of birth, we stratified our analysis into three subgroups: nulliparous, parous without previous caesarean and parous with previous caesarean. As health problems before or during pregnancy and a health care provider trying to induce labour may be on the causal pathway between BMI, GWG and mode of birth, we assessed results from models with and without these variables.

The contribution to caesarean births of overweight or obese BMI and more than recommended GWG was estimated using population attributable fractions (PAFs). The calculation of PAFs has the advantage of incorporating the increased risk due to high BMI or GWG and the prevalence of these two determinants, in order to provide an estimate of the potential reduction in caesarean birth if high BMI and GWG were eliminated. We calculated PAFs directly from our multivariable logistic regression models using the sequential and average attributable fraction method which takes into account that ORs are adjusted for confounders [22].

All analyses were carried out using sampling weights. We calculated variances using bootstrap weights to capture the variability introduced by the sample design and weighting adjustments [23]. We used SAS EG software, Version 5.1, copyright SAS Institute Inc. [24]. Review by an ethics board was not required, as the MES data are anonymous and this study did not generate identifying information.

The overall caesarean birth rate was 25.7%, with substantial variation across parity and prior caesarean group strata. Among nulliparous women, parous women with no prior caesarean and parous women with a prior caesarean, caesarean birth rates were 29.6%, 5.8% and 80.2% respectively. Among nulliparous and parous women without a previous caesarean, rates in obese women were 45.1% and 9.7% respectively, and rates in women who gained above their recommended GWG were 33.5% and 8.0% respectively (Table 4). The incidence of caesarean births increased in all groups as BMI and GWG increased, except for GWG among parous women with a prior caesarean. In this group the trend was reversed; but adjusted ORs were not significantly different (Table 4). The high caesarean birth rates (above 75%) in parous women with a prior caesarean and low rates (below 10%) in parous women with no prior caesarean limited the scope for detecting significant decreases or increases in risk in these groups.

The adjusted risk of caesarean birth did not differ significantly between underweight and normal-weight women; it also did not differ significantly between women who gained less than the recommended amount and those who gained within the recommended amount (Table 4). The risk was significantly elevated among women who were overweight (OR = 1.23 [1.04-1.47]), obese (OR = 1.95 [1.61-2.36]), or had gained more than the recommended amount (OR = 1.36 [1.17-1.59]). Among nulliparous women who were overweight, obese or above their recommended GWG, the overall risk of caesarean birth was increased, with most of the increase attributable to unplanned caesareans. Among parous women with no prior caesarean, the risk of caesarean birth was low, but significantly elevated among those who were obese (OR = 2.03 [1.17-3.53]) or above their recommended GWG (OR = 1.75 [1.13-2.71]). There was no significant relationship between caesarean births and BMI or GWG among women with a previous caesarean (Table 4). Adjusting for women’s reports that their health care provider tried to induce labour, or that they experienced health problems before or during pregnancy, did not significantly change these risk patterns (data not shown).

The fractions of caesarean births associated with overweight or obese BMI and more than recommended GWG are presented in Table 5. Among all women, 10.1% (9.9-10.2) of caesareans were associated with overweight or obese BMI and 10.9% (10.7-11.1) were associated with above recommended GWG. One in five caesareans (20.2% [20.0-20.4]) was associated with either above normal BMI or excess GWG. Results were similar for nulliparous women. In parous women with no previous caesarean, the proportion of caesareans associated with above recommended GWG was twice that of overweight or obese BMI (23.6% [23.0-24.2] versus 10.9% [10.4-11.4]). Almost one third (31.8% [31.3-32.4]) of caesarean births in this group were associated with either above normal BMI or excess GWG.