Background
Nonalcoholic fatty liver disease causes hepatic insulin resistance and is associated with metabolic syndrome. Elevated levels of alanine aminotransferase are associated with nonalcoholic fatty liver disease. The effect of hepatic insulin resistance is not only increased glycogen breakdown but also liberation of free fatty acids due to increased lipolysis. Both of these fuel sources are associated with macrosomia. There is little known about the impact of maternal nonalcoholic fatty liver disease on excessive fetal growth.
Objective
The purpose of this study was to investigate the association of early elevated alanine aminotransferase with large-for-gestational-age birthweight.
Study Design
This is a secondary analysis from a nested case-control study of maternal alanine aminotransferase values measured between 8–18 weeks and subsequent gestational diabetes. We included women with singleton gestations with complete delivery information and without known diabetes, liver disease, or moderate self-reported alcohol use during pregnancy. We used inverse probability weighting to standardize the population and minimize selection bias. We calculated population-based birthweight z scores and defined large for gestational age as ≥90th percentile for gestational age. We compared maternal baseline characteristics with analysis of variance, Fisher exact test, or Wilcoxon rank sum. We then performed conditional logistic regression to evaluate the relationship between alanine aminotransferase and large for gestational age adjusting for maternal age, body mass index, parity, gestational diabetes, smoking, and maternal weight gain.
Results
We identified 26 cases of large for gestational age out of 323 mother-infant dyads. The mean maternal body mass index was higher in the large-for-gestational-age group compared to controls (33.7 [SD 4.3] vs 28.9 [SD 6.5], P = .002). Large-for-gestational-age babies were less likely to be male (8 [31%] vs 172 [58%], P = .01) and had a higher mean gestational age (39.5 [SD 0.9] vs 38.4 [SD 2.3] weeks, P = .01). Maternal and infant characteristics were otherwise similar. The mean alanine aminotransferase among the large-for-gestational-age cases was 28 (SD 37) U/L compared to 16 (SD 8) U/L for controls. Each unit increase in log-transformed alanine aminotransferase was associated with a 3-fold odds of large for gestational age (adjusted odds ratio, 3.05; 95% confidence interval, 2.27–4.10; P < .0001), and alanine aminotransferase ≥90th percentile (26 U/L) was associated with a 4-fold increased odds of large for gestational age (adjusted odds ratio, 4.03; 95% confidence interval, 2.84–5.70; P < .0001). This association was unchanged when analysis was restricted only to women without gestational diabetes with a glucose loading test <120 mg/dL (log-transformed alanine aminotransferase: adjusted odds ratio, 3.05; 95% confidence interval, 1.04–8.96; P = .04, and alanine aminotransferase ≥90th percentile: adjusted odds ratio, 4.21; 95% confidence interval, 1.20–14.82; P = .03).
Conclusion
Unexplained elevated alanine aminotransferase in the first trimester was associated with a 4-fold increase in the odds of large-for-gestational-age birthweight even in the absence of clinical glucose intolerance. This may represent the impact of maternal nonalcoholic fatty liver on the fetal developmental milieu.
Introduction
Pregestational and gestational diabetes are well-described risk factors for excessive fetal growth. Recent analyses have identified prepregnancy body mass index (BMI) and gestational weight gain as additional independent predictors of fetal macrosomia. Small studies have found associations with various maternal metabolic factors and large-for-gestational-age (LGA) infant birthweight, including maternal triglycerides, free fatty acids (FFA), and leptin. While many elements of metabolic syndrome have been investigated as possible risk factors for fetal overgrowth, there has been little investigation dedicated to the impact of nonalcoholic fatty liver on birthweight.
Nonalcoholic fatty liver disease (NAFLD) is considered to be the hepatic component of metabolic syndrome, affecting an estimated 10–35% of adults in the United States. In NAFLD, patients have an imbalance of fatty acid uptake and de novo lipogenesis and fatty acid output resulting in hepatic infiltration that is associated with hepatic insulin resistance. It is associated with concurrent and incident type 2 diabetes mellitus (DM). NAFLD is a pathologically defined diagnosis, but is frequently first identified in a clinical setting following unexplained elevated alanine aminotransferase (ALT). In the absence of viral hepatitis, hepatobiliary disease, or alcohol abuse, isolated elevated ALT is strongly correlated with insulin resistance, independent of other risk factors and in a manner that increases with escalating BMI. Gómez-Sámano et al demonstrated that among subjects with metabolic syndrome or other impaired insulin sensitivity, including DM, incremental increase in ALT throughout the normal range is associated with increased hepatic insulin resistance index. In normal pregnancies, ALT does not significantly vary across trimesters and values are commonly lower than prepregnancy values. Nothing is known about the pattern of ALT across trimesters in women with NAFLD.
Based on the established relationship between unexplained elevated ALT and hepatic insulin resistance, we hypothesized that maternal ALT in early pregnancy would be associated with LGA birthweight. We therefore sought to identify the relationship between neonatal birthweight and maternal ALT measured early in pregnancy.
Materials and Methods
Study population
This was a planned secondary analysis of a nested case-control cohort designed to explore the relationship between early ALT and subsequent diagnosis of gestational diabetes. Cases of gestational DM (GDM) and BMI-matched controls were selected from a cohort of women recruited from Brigham and Women’s hospital, an urban academic medical center in Boston, MA, from 2006 through 2013 as part of a prospective longitudinal study of biomarkers and pregnancy outcomes. Subjects were excluded if they had multiple gestation, known DM type 1 or 2, prepregnancy alcoholic abuse and/or moderate alcohol use during pregnancy by self-report, or known viral or immunologic liver disease. In brief, we identified 83 women with gestational diabetes diagnosed either by 3-hour glucose tolerance test meeting Carpenter-Coustan criteria, or women diagnosed clinically who required insulin in the third trimester for glycemic control and 247 women used as controls in the primary analysis who had a glucose loading test (GLT) <120 mg. Seven subjects did not have birthweights available due to loss to follow-up (1 GDM case and 6 controls). Among the remaining cohort, we identified infants with LGA birthweight and compared ALT values in those mothers to those of non-LGA infants. In the absence of prior data in this field to inform a power estimate, we used all available samples. The study was approved by the institutional review board at Brigham and Women’s Hospital (protocol no. 2009P000810).
Exposure assessment
Plasma samples were collected at a mean gestational age of 10.6 (SD 2.6) weeks and stored at –80°C. ALT was assayed using the ARCHITECT c System (Abbot Diagnostics, Abbott Park, IL) performed at the Massachusetts General Hospital clinical laboratories. The assay has a lower limit of detection of 5.1 U/L with a range up to 4113 U/L.
Outcome assessment
Birthweights were extracted from the medical record by trained reviewers. We defined LGA as birthweight >90th percentile for completed week of gestational age (z score >1.28) by published sex-specific growth curves.
Covariate assessment
Trained reviewers abstracted medical records to obtain relevant pregnancy outcomes. Demographic (including maternal age, race, and ethnicity) and historical (including pregnancy history, relevant medical history, and medication use) information was collected by participant questionnaire at study enrollment. Maternal BMI was calculated using reported prepregnancy weight and height measured in the first trimester.
Statistical analysis
For all analyses we normalized the population using inverse probability weighting to correct the overrepresentation of gestational diabetes prior to this analysis, a well-validated method of standardizing a nested case control. This adjustment negates the effect of oversampling gestational diabetes and makes results generalizable to pregnant women in the base patient cohort. Statistical testing of ALT results was performed after logarithmic transformation to normalize the distribution. We compared baseline sociodemographics and prevalence of selected pregnancy characteristics of cases and controls using Wilcoxon rank sum test for continuous variables and χ 2 , Fisher exact, or analysis of variance for categorical variables where appropriate. In fully adjusted multivariate conditional logistic regression models, maternal age and gestational diabetes were included a priori, and additional covariates were added in a forward stepwise model selection procedure with inclusion in final models if they altered effect estimates by >10%. Variables that were considered included health insurance category, smoking status, use of artificial reproductive technology, total maternal weight gain, parity, and gender of infant as well as history of thyroid disease, history of seizure disorders, and use of pain medication as factors that could affect ALT. As it was an exploratory analysis, we evaluated both a continuous association of LGA with log-transformed ALT as well as association of LGA with a preselected cutoff of the subject population 90th percentile. All analyses were performed with software (SAS, Version 9.4; SAS Institute Inc, Cary, NC). P levels < .05 were used to define statistical significance.
Materials and Methods
Study population
This was a planned secondary analysis of a nested case-control cohort designed to explore the relationship between early ALT and subsequent diagnosis of gestational diabetes. Cases of gestational DM (GDM) and BMI-matched controls were selected from a cohort of women recruited from Brigham and Women’s hospital, an urban academic medical center in Boston, MA, from 2006 through 2013 as part of a prospective longitudinal study of biomarkers and pregnancy outcomes. Subjects were excluded if they had multiple gestation, known DM type 1 or 2, prepregnancy alcoholic abuse and/or moderate alcohol use during pregnancy by self-report, or known viral or immunologic liver disease. In brief, we identified 83 women with gestational diabetes diagnosed either by 3-hour glucose tolerance test meeting Carpenter-Coustan criteria, or women diagnosed clinically who required insulin in the third trimester for glycemic control and 247 women used as controls in the primary analysis who had a glucose loading test (GLT) <120 mg. Seven subjects did not have birthweights available due to loss to follow-up (1 GDM case and 6 controls). Among the remaining cohort, we identified infants with LGA birthweight and compared ALT values in those mothers to those of non-LGA infants. In the absence of prior data in this field to inform a power estimate, we used all available samples. The study was approved by the institutional review board at Brigham and Women’s Hospital (protocol no. 2009P000810).
Exposure assessment
Plasma samples were collected at a mean gestational age of 10.6 (SD 2.6) weeks and stored at –80°C. ALT was assayed using the ARCHITECT c System (Abbot Diagnostics, Abbott Park, IL) performed at the Massachusetts General Hospital clinical laboratories. The assay has a lower limit of detection of 5.1 U/L with a range up to 4113 U/L.
Outcome assessment
Birthweights were extracted from the medical record by trained reviewers. We defined LGA as birthweight >90th percentile for completed week of gestational age (z score >1.28) by published sex-specific growth curves.
Covariate assessment
Trained reviewers abstracted medical records to obtain relevant pregnancy outcomes. Demographic (including maternal age, race, and ethnicity) and historical (including pregnancy history, relevant medical history, and medication use) information was collected by participant questionnaire at study enrollment. Maternal BMI was calculated using reported prepregnancy weight and height measured in the first trimester.
Statistical analysis
For all analyses we normalized the population using inverse probability weighting to correct the overrepresentation of gestational diabetes prior to this analysis, a well-validated method of standardizing a nested case control. This adjustment negates the effect of oversampling gestational diabetes and makes results generalizable to pregnant women in the base patient cohort. Statistical testing of ALT results was performed after logarithmic transformation to normalize the distribution. We compared baseline sociodemographics and prevalence of selected pregnancy characteristics of cases and controls using Wilcoxon rank sum test for continuous variables and χ 2 , Fisher exact, or analysis of variance for categorical variables where appropriate. In fully adjusted multivariate conditional logistic regression models, maternal age and gestational diabetes were included a priori, and additional covariates were added in a forward stepwise model selection procedure with inclusion in final models if they altered effect estimates by >10%. Variables that were considered included health insurance category, smoking status, use of artificial reproductive technology, total maternal weight gain, parity, and gender of infant as well as history of thyroid disease, history of seizure disorders, and use of pain medication as factors that could affect ALT. As it was an exploratory analysis, we evaluated both a continuous association of LGA with log-transformed ALT as well as association of LGA with a preselected cutoff of the subject population 90th percentile. All analyses were performed with software (SAS, Version 9.4; SAS Institute Inc, Cary, NC). P levels < .05 were used to define statistical significance.
Results
Demographic and clinical characteristics of the study population are presented in Table 1 . We identified 26 (7%) LGA infants among the 323 mother-infant dyads. Twenty LGA infants were in the population with normal GLT and 6 LGA infants were delivered by women with GDM. The mean maternal BMI was higher in the LGA group compared to non-LGA (33.7 [SD 4.3] vs 28.9 [SD 6.5]; P = .002) and LGA babies were less likely to be male (31% vs 58%; P = .01) and had a higher mean gestational age at delivery (39.5 [SD 0.9] vs 38.4 [SD 2.3] weeks; P = .01) ( Table 1 ). Maternal and infant characteristics were otherwise similar.
