An estimated four to twelve percent of pregnancies are complicated by some degree of abnormal glucose intolerance (AGT), including gestational diabetes mellitus (GDM) as well as milder degrees of glucose tolerance [¹]; rates are higher in Hispanic women as compared to non-Hispanic white women. Recently, the 7-year prospective Hyperglycemia and Adverse Pregnancy Outcome study found a consistent, linear increase in risk of adverse perinatal outcomes (cesarean section, fetal size, neonatal hypoglycemia, and fetal hyperinsulinemia) over the entire range of maternal blood glucose levels [²], not just with GDM, suggesting that identifying risk factors for milder degrees of disturbances in glucose metabolism is critical for maternal and offspring health.

Cigarette smoking remains common during pregnancy [³,⁴], but research examining the role of smoking as a risk factor for AGT is limited. Smoking has been associated with increased insulin resistance and type 2 diabetes in non-pregnant women [⁵]. However, whether smoking is a risk factor for glucose disturbances during pregnancy is less clear. Physiologic studies suggest that smoking may increase circulating levels of catecholamines, compounds involved in insulin resistance, and that nicotine may have direct toxic effects on pancreatic beta cell function [^6-8].

Epidemiologic evidence supporting the role of smoking as a risk factor for GDM is limited and there is major heterogeneity in research design [⁹]. With few exceptions, the majority of studies have examined smoking as a dichotomous variable [^10-15] and only five studies presented adjusted odds ratios [^16-20]. Furthermore, only one prior study has been conducted in a Hispanic population [²¹]. There is some evidence that nicotine metabolism varies by race, leading to higher and lower nicotine exposure per cigarette among African-Americans and Chinese-Americans, respectively, although Hispanics appear to metabolize nicotine at the same rate as whites [²²,²³]. In addition, smoking patterns vary markedly among US Hispanic women according to country of origin, with Puerto Rican women having the highest rates of smoking during pregnancy compared to Hispanics from other countries [⁴]. One study found that 41% of female Puerto Rican smokers preferred menthol brands of cigarettes, and that menthol smokers may have larger inhaled volumes and higher cotinine levels than non-menthol smokers [²⁴]. These characteristics of cigarette use among Hispanics may impact the effect of smoking on AGT in this population.

This is important as Hispanics are one of the fastest-growing segments of the US population, comprising 14.7% of the female population in 2008 [²⁵]. Over 24% of all US births in 2007 were to Hispanics [²⁶], representing the highest fertility rate of any race or ethnic group. Health status varies widely between Hispanic subgroups, with rates of adverse birth outcomes highest for Puerto Rican women [²⁷].

Therefore, we prospectively examined the association between smoking during pre, early, and mid pregnancy and risk of AGT among a population of predominantly Puerto Rican prenatal care patients at a large tertiary care facility. We hypothesized that smokers during pre, early, or mid pregnancy would have an increased risk of AGT as compared to non-smokers. The current study contributes to the existing literature by examining the association between smoking during pregnancy, including cigarettes per day, and risk of AGT in a high-risk population of Hispanic prenatal care patients. To our knowledge, this is first prospective study of smoking and AGT in a Puerto Rican population.

The majority of women were young (72% > 25 years), 55% had not graduated from high school and 60% had an annual household income less than $15,000 (Table 1). The population had a number of risk factors for AGT including overweight or obesity (49%) and family history of diabetes (64%).

A total of 119 (11.8%) participants were classified with AGT. Higher levels of age, education, income, pre-pregnancy BMI, family history of type 2 diabetes, and personal history of GDM were associated with increased risk of AGT (Table 1). In addition, women with AGT were more likely to have hypertension in the current pregnancy. Parity, weight gain up to the time of GDM screen, language preference, birthplace, alcohol consumption and pregnancy physical activity were not associated with AGT.

Approximately one-third of participants (35%) reported smoking cigarettes prior to pregnancy, while 20% of women reported smoking cigarettes during early pregnancy and 17% in mid pregnancy (Table 2). Women who smoked during pregnancy were predominantly light smokers, with 16% smoking > 0-9 cigarettes per day and 4% smoking 10 or more cigarettes per day in early pregnancy. Compared to non-smokers, smokers were more likely to be multiparous, have lower education level and household income, and to drink alcohol during pregnancy (data not shown). Smokers were also more likely to prefer using only English and to be born in the United States.

After adjusting for age and early pregnancy smoking, women who smoked in pre-pregnancy did not have an increased risk of AGT (OR = 1.42; 95% CI 0.82-2.46) as compared to non-smokers. However, when evaluated according to dose, women who smoked > 0-9 cigarettes per day in pre-pregnancy had an increased risk of AGT (OR = 1.90; 95% 1.02-3.55) compared to pre-pregnancy non-smokers, while those who smoked 10+ cigarettes per day were not at increased risk (OR = 0.95; 95% 0.45-2.00) (Table 2). The increased risk for light smoking during pre-pregnancy was attenuated after adjustment for gestational weight gain (OR = 1.73; 95% CI 0.84-3.58). Similarly, smokers in early (OR = 0.48; 95% CI 0.21, 1.10) and mid pregnancy (OR = 0.38; 95% CI 0.13-1.11) did not have an increased risk of AGT as compared to non-smokers in multivariable models. When we restricted the analysis to participants with complete smoking data in all three time periods (n = 609) findings were virtually unchanged.

We then examined the association between smoking category and continuous glucose levels from the 1-hr OGTT (mean = 101.9 mg/dL) (Table 3). Regardless of dose, pre-pregnancy smoking was not associated with glucose level in age-adjusted and multivariable (β = -2.00, p = 0.36) models. However, women who smoked in early pregnancy (β = -4.60, p = 0.03) and in mid-pregnancy (β = -6.08, p = 0.01) had lower mean glucose levels as compared to non-smokers after adjusting for AGT risk factors.

Finally, to determine if the association between smoking and risk of AGT was confounded by dietary intake, we evaluated the association between smoking during pregnancy and AGT in the subset of women with complete dietary intake information (n = 566). Smokers consumed significantly more total fat (p = 0.004) and saturated fat (p = 0.0003) and significantly less fiber (p = 0.005), carbohydrates (p = 0.01), and fewer calories (p = 0.05) than non-smokers. After adjusting for dietary factors, findings were not substantively changed. For example, the multivariable OR for early pregnancy smokers as compared to non-smokers was 0.48 (95% CI 0.18, 1.30).

Women who were not screened for GDM (n = 54) did not differ from those who were screened for GDM in terms of age, parity, education, BMI, family history of diabetes, history of GDM, language preference, or smoking in pre or early pregnancy (all p > 0.20). Similarly, women missing smoking data during any pregnancy time period did not differ from women missing such data during other pregnancy time periods, nor from women with complete information on smoking.

In this prospective cohort study of Hispanic women, results do not suggest an increased risk of AGT related to smoking cigarettes. There was the suggestion that light smoking in pre-pregnancy was associated with an increased risk of AGT compared to non-smokers. Smoking during early and mid pregnancy was not associated with AGT. Cigarette smoking during early and mid pregnancy were associated with lower mean screening glucose values on the 1-hr OGTT glucose tolerance test as compared to non-smokers, independent of weight gain during pregnancy.

Our findings of a suggestion of an increased risk for pre-pregnancy light smoking is consistent with findings among 14,613 participants in the Nurses Health Study [³³]. Specifically, Solomon et al. found that, compared to non-smokers, pregravid smokers of 5+ cigarettes per day had an increased risk of GDM (RR = 1.65; 95% CI 1.05-2.58) compared to pregravid non-smokers. It is possible, however, that observed increased risks among pre-pregnancy smokers may be due to the effect of quitting smoking. Quitting or reducing the number of cigarettes smoked during pregnancy may lead to increased gestational weight gain [³⁴], a risk factor for AGT and GDM [³⁵]. Indeed, after controlling for gestational weight gain, our observed increased risk for pre-pregnancy smoking was attenuated.

Similarly, we observed that smoking in early and mid pregnancy were not associated with AGT in age-adjusted or multivariable analyses. Prior epidemiologic studies have been conflicting. In a recent review article, Wendland et al. obtained a crude odds ratio of 1.03 (99% CI 0.85-1.25) and an adjusted odds ratio of 0.95 (95% CI 0.85-1.07) for smoking during pregnancy and risk of GDM [⁹]. In the only prior study to assess Hispanic women (Brazil), Wendland et al. found that women who smoked during pregnancy had a reduction in risk of GDM (OR = 0.69; 95% CI 0.50-0.96) after adjusting for GDM risk factors [²¹]. While there were no significant differences in fasting and 1 h plasma glucose means, those for 2 h glucose were significantly lower in smokers (5.58+0.04 mmol/L), p < 0.01) in adjusted analyses. Similarly, we found that smokers in early and mid pregnancy had lower levels of plasma screening glucose values as compared to non-smokers.

There is biologic plausibility for such an association. Nicotine has been shown to block the release of inflammatory cytokines, inhibiting inflammation associated with insulin resistance and thereby diminishing hyperglycemia [⁹]. Smoking-induced release of catecholamines may also increase the metabolism of soluble nutrients, thereby lowering glucose levels among smokers [³⁶]. Smoking has also been consistently and negatively associated with pre-eclampsia via physiological mechanisms that may be similar [³⁷]. These observations, however, do not exclude the fact that smoking may increase risk of type 2 diabetes in non pregnant women through different metabolic pathways. In addition, the higher doses of smoking during the pre- or non-pregnancy periods [³⁸] may explain, in part, differences in associations between pre-pregnancy vs. pregnancy smoking and risk of AGT. Finally, it is important to note the established adverse effects of smoking on birth outcomes.

Timing of smoking measurement during pregnancy may also explain, in part, differences in study findings. This meta analysis found that studies that assessed smoking earlier in pregnancy [¹¹,^13-15,¹⁸] showed a tendency to present an increased risk of GDM, while studies that ascertained smoking later [¹⁰,¹⁶,¹⁹] showed odds ratios slightly less than 1. The authors attribute elevated risks observed in early pregnancy to failure to identify smokers who quit or reduced the number of cigarettes smoked, therefore erroneously attributing the association found to smoking rather than to its reduction and subsequent possible increase in weight gain. Smoking assessed later in pregnancy may therefore be a more accurate assessment of true smoking status in pregnancy.

Our study had several limitations. The relatively small number of cases of AGT likely limited our power to detect associations in analyses of AGT as a dichotomous outcome and also precluded evaluating risk of frank GDM. However, this limitation was not faced by the analysis of screening glucose level as a continuous outcome. Like all studies relying on self-reported smoking information, there may be some degree of misclassification in the smoking variable. However, while women may be more susceptible to biased reporting in smoking cessation trials [³⁹], evidence suggests that women are truthful when self-reporting this information in observational studies [^40-43]. A study by Christensen et al. [⁴²] also using a structured, interviewer-administered questionnaire found high validity for self-reported smoking when compared to exhaled carbon monoxide levels. Ross et al. [⁴⁰] compared self-reported cigarette smoking with serum cotinine levels among pregnant women in Minnesota and found that 96% of women who denied smoking reported their status correctly (kappa = 0.86). The validity of self-reported smoking during pregnancy has not been studied in Hispanic women. However, the relatively high smoking rates observed in the current study (35% pre-pregnancy, 20% early pregnancy) suggest that underreporting of smoking may be less of a concern. In addition, by incorporating a category for former smokers in our analysis, we avoid a limitation faced by prior studies that combined ex-smokers with never smokers [⁹].

Although we had extensive information on established risk factors for AGT, dietary variables were only available for a subset of women who completed the FFQ at the mid pregnancy interview. However, findings in this group were consistent with overall study findings. Similarly, women who were not screened for GDM did not differ from screened women in terms of any demographic, health, or smoking variables. Women who were missing information on smoking did not differ from women who were not missing this information.

Participants in this study were healthy Hispanic women with a singleton pregnancy receiving prenatal care. Because participants were predominantly light smokers, we were not able to evaluate the effect of heavy smoking on AGT. There is also some evidence that nicotine metabolism varies by race [²²,²³]. Therefore, care should be taken in generalizing to other racial or ethnic groups whose smoking and AGT patterns may differ. Indeed, the cigarette brands, tar content, inhalation behavior, age of first use, and overall patterns of smoking among Hispanics may differ from other populations [²⁴,²⁸], and these characteristics of cigarette use among Hispanics may impact the effect of smoking on AGT.

Finally, BMI, parity, age, and gestational weight gain may be important effect modifiers of the association between smoking and AGT [⁹]. However, due to our limited sample size we did not have adequate power to assess effect modification by these factors with sufficient precision. Future studies in Puerto Rican women should evaluate the effect of these variables.

To our knowledge this is the first study to examine the association between smoking and AGT among a Puerto Rican population. This study contributes to the sparse literature on smoking and AGT. We did not observe any increased risk of AGT associated with smoking during pregnancy; findings should be confirmed in other study populations.

The authors declare that they have no competing interests.

AH conceived of the study question, analyzed the data and drafted the manuscript. EB, PP, and EC assisted with data analysis, interpretation, and manuscript editing. RT assisted with data analysis. LC designed the study and assisted with data analysis, interpretation and manuscript editing. All authors read and approved the final manuscript.

The pre-publication history for this paper can be accessed here:

http://www.biomedcentral.com/1471-2393/10/55/prepub

Financial support: American Diabetes Association Career Development Award 7-00-CD-02.