Directed acyclic graphs: a tool for causal studies in paediatrics

Many paediatric clinical research studies, whether observational or interventional, have as an eventual aim the identification or quantification of causal relationships. One might ask: does screen time influence childhood obesity? Could overuse of paracetamol in infancy cause wheeze? How does breastfeeding affect later cognitive outcomes? In this review, we present causal directed acyclic graphs (DAGs) to a paediatric audience. DAGs are a graphical tool which provide a way to visually represent and better understand the key concepts of exposure, outcome, causation, confounding, and bias. We use clinical examples, including those outlined above, framed in the language of DAGs, to demonstrate their potential applications. We show how DAGs can be most useful in identifying confounding and sources of bias, demonstrating inappropriate statistical adjustments for presumed biases, and understanding threats to validity in randomised controlled trials. We believe that a familiarity with DAGs, and the concepts underlying them, will be of benefit both to the researchers planning studies, and practising clinicians interpreting them.

Hypothermia at neonatal intensive care unit admission was not associated with respiratory disease or death in very preterm infants

AIM

This study investigated the association between hypothermia and respiratory distress syndrome (RDS), bronchopulmonary dysplasia (BPD) or death in very preterm infants admitted to a Danish neonatal intensive care unit (NICU).

METHODS

We studied 675 infants born at Aalborg University Hospital before 32 weeks and admitted to the NICU from April 1997 to December 2011. Hypothermia was defined as a core temperature of <36.5°C on admission. The primary outcome was severe RDS or death within the first three days of life, and the secondary outcome was BPD or death before 36 postmenstrual weeks. The multivariable logistic regression was adjusted for early-onset infection, gestational age, Apgar score, sex, treatment year and birth weight.

RESULTS

Infants with hypothermia had a twofold increase (OR) in the odds for RDS or death (2.03), but the adjusted OR was not statistically significant (1.36). They also demonstrated a twofold increase (OR) in the odds for BPD or death (2.28), but again the adjusted OR was not statistically significant (1.03).

CONCLUSION

After adjusting for confounders, we found that the association between hypothermia on admission to the NICU and RDS or death, or BPD or death was statistically insignificant.

Congenital Heart Defects and Indices of Placental and Fetal Growth in a Nationwide Study of 924 422 Liveborn Infants

BACKGROUND

Congenital heart defects (CHDs) have been associated with placental anomalies. The nature and the consequences of this association remain poorly understood. We aimed to estimate the associations between all major subtypes of CHD and placental weight at birth, and the association between placental weight and measures of both overall and cerebral growth in fetuses with CHD, as well.

METHODS

We included all 924 422 liveborn Danish singletons, 1997 to 2011. CHD was present in 7569. We compared mean differences in placental weight z score between newborns with CHD and newborns without CHD by multivariable linear regression adjusted for potential confounders.

RESULTS

CHD was associated with a mean z score difference of -0.04 (95% confidence interval, -0.07 to -0.02). Some subtypes were associated with smaller placental size at birth: tetralogy of Fallot, -0.45 (95% confidence interval, -0.58 to -0.31); double-outlet right ventricle, -0.48 (95% confidence interval, -0.87 to -0.10); major ventricular septal defects, -0.41 (95% confidence interval, -0.52 to -0.29). Placental weight z score was associated with birth weight and head circumference z scores in all subtypes. In the 3 mentioned subtypes, the mean deviations from the population mean head circumference and birth weight z scores were reduced by up to 66% with adjustment for placental weight z score.

CONCLUSIONS

Three subtypes of CHD were associated with lower placental weight, and placental weight was associated with measures of both overall growth and cerebral growth in fetuses with all subtypes of CHD. In certain subtypes, the described deviations in fetal growth were reduced by up to two-thirds after adjustment for placental weight z score.

Congenital Heart Defects and Measures of Fetal Growth in Newborns with Down Syndrome or 22q11.2 Deletion Syndrome

OBJECTIVES

To estimate the association between congenital heart defects (CHD) and indices of fetal growth in Down and 22q11.2 deletion syndromes.

STUDY DESIGN

We established 2 Danish nationwide cohorts of newborn singletons with either Down syndrome (n = 670) or 22q11.2 deletion syndrome (n = 155), born 1997-2011. In both cohorts, we analyzed the association between CHD, CHD severity, and indices of fetal growth by multivariable linear regression adjusted for potential confounders. We report mean differences in gestational age specific z-scores compared with newborns without CHD.

RESULTS

Down syndrome and 22q11.2 deletion syndrome were both associated with lower mean birth weight and head circumference z-scores. We found no association between CHD or CHD severity and indices of fetal growth. In Down syndrome, the association between any CHD and the mean difference in head circumference z-score was 0.03 (95% CI -0.12, 0.18), and the estimate regarding birth weight z-score was 0.09 (95% CI -0.08, 0.25). The corresponding estimates in 22q11.2 deletion syndrome were 0.00 (95% CI -0.33, 0.32) and -0.09 (95% CI -0.45, 0.26).

CONCLUSIONS

We found no association between CHD and fetal growth measures in newborns with Down syndrome or 22q11.2 deletion syndrome. Thus, in certain subtypes of CHD, the contribution of genetic factors to prenatal growth impairment may be more important than circulatory disturbances.

Determinants of serum levels of perfluorinated alkyl acids in Danish pregnant women

Humans are exposed to perfluorinated alkyl acids (PFAAs) from food, drinking water, air, dust, and consumer products. PFAAs are persistent and bio-accumulative. In the present study, we aimed to establish how the serum levels of PFAAs differ according to age, pre-pregnancy body mass index (BMI), previous miscarriages, educational level, country of birth, smoking, and alcohol intake. We included 1438 Danish pregnant nulliparous women from the Aarhus Birth Cohort. The women gave a blood serum sample between week 11 and 13 of pregnancy. Sixteen PFAAs were extracted from serum using solid phase extraction and analyzed by liquid chromatography/tandem mass spectrometry. Multivariable linear regression analysis was used to determine the associations between individual characteristics of the women and their levels of seven PFAAs that were detected in at least 50% of the samples. The total concentration of the PFAAs (∑PFAA) was higher in older women. On average, normal weight women had a higher ∑PFAA level than underweight, overweight, and obese women. Higher levels were also observed for women without previous miscarriages, women with a high educational level, women born in Denmark (as opposed to women born elsewhere but currently living in Denmark), non-smokers, and women who consumed alcohol before or during pregnancy. These associations were similar for all the studied PFAAs, although the levels of perfluoroundecanoic acid varied more across the categories of age, BMI, education, smoking, and alcohol consumption than any other PFAAs measured.

Time trends of perfluorinated alkyl acids in serum from Danish pregnant women 2008-2013

We aimed to estimate the levels and time trends of perfluorinated alkyl acids (PFAAs) in serum of 1533 Danish pregnant nulliparous women between 2008 and 2013. The selection criterion of only including nulliparous women was chosen to avoid confounding from parity. The serum samples were analyzed for sixteen PFAAs using solid phase extraction and liquid chromatography tandem mass spectrometry (LC-MS/MS). We investigated the time trends for seven PFAAs, which were detected in more than 50% of the samples: perfluorohexane sulfonate (PFHxS), perfluoroheptane sulfonate (PFHpS), perfluorooctane sulfonate (PFOS), perfluorooctanoic acid (PFOA), perfluorononanoic acid (PFNA), perfluorodecanoic acid (PFDA), and perfluoroundecanoic acid (PFUnA). We found that the serum levels of all seven PFAAs decreased during the period from 2008 to 2013; on average PFHxS decreased with 7.0% per year, PFHpS with 14.8%, PFOS with 9.3%, PFOA with 9.1%, PFNA with 6.2%, PFDA with 6.3%, and PFUnA with 7.1% per year. Adjustment for maternal age, body mass index (BMI), educational level and gestational age at blood sampling did not change the time trends much. To our knowledge, we are the first to report decreasing trends of PFNA, PFDA and PFUnA since year 2000, thereby indicating that the phase-out of these compounds are beginning to show an effect on human serum levels.

Congenital Heart Defects and Indices of Fetal Cerebral Growth in a Nationwide Cohort of 924 422 Liveborn Infants

BACKGROUND

Neurodevelopmental disorders are the most common and distressful comorbidities associated with congenital heart defects (CHD). Head circumference at birth (HC), a proxy for prenatal cerebral growth, is an established risk factor for neurodevelopmental disorders.

METHODS AND RESULTS

In a nationwide cohort, we included all 924 422 liveborn Danish singletons, 1997 to 2011. CHD was present in 5519. The association between CHD and growth indices was analyzed by multivariable linear regression, adjusted for potential confounders. We report mean differences in gestational age-specific z scores in comparison with the general population. CHD was associated with lower HC z scores, -0.10 (95% confidence interval [CI], -0.13 to -0.08). Several CHD subtypes were associated with smaller HC, eg, hypoplastic left heart syndrome, -0.39 (95% CI, -0.58 to -0.21); common arterial trunk, -0.41 (95% CI, -0.74 to -0.09); and major ventricular septal defects, -0.25 (95% CI, -0.35 to -0.15). Other single-ventricle defects, transposition of the great arteries, tetralogy of Fallot, and anomalous pulmonary venous return, were also associated with smaller HC. Transposition of the great arteries was associated with smaller HC relative to birth weight, -0.26 (95% CI, -0.39 to -0.13). Major ventricular septal defects were associated with larger HC relative to birth weight. The results were consistent under various conditions, eg, when siblings of infants with CHD (n=5311) or infants with other major malformations (n=24 974) were used as the reference.

CONCLUSIONS

Several subtypes of CHD were associated with smaller HC. The associations with major ventricular septal defects, common arterial trunk, and anomalous pulmonary venous return have not previously been described. Only infants with transposition of the great arteries had smaller HC relative to birth weight.

Perfluoroalkyl substances and time to pregnancy in couples from Greenland, Poland and Ukraine

BACKGROUND

Perfluoroalkyl substances (PFAS) are suggested to affect human fecundity through longer time to pregnancy (TTP). We studied the relationship between four abundant PFAS and TTP in pregnant women from Greenland, Poland and Ukraine representing varying PFAS exposures and pregnancy planning behaviors.

METHODS

We measured serum levels of perfluorooctanoic acid (PFOA), perfluorooctane sulfonate (PFOS), perfluorohexane sulfonic acid (PFHxS) and perfluorononanoic acid (PFNA) in 938 women from Greenland (448 women), Poland (203 women) and Ukraine (287 women). PFAS exposure was assessed on a continuous logarithm transformed scale and in country-specific tertiles. We used Cox discrete-time models and logistic regression to estimate fecundability ratios (FRs) and infertility (TTP >13 months) odds ratios (ORs), respectively, and 95% confidence intervals (CI) according to PFAS levels. Adjusted analyses of the association between PFAS and TTP were done for each study population and in a pooled sample.

RESULTS

Higher PFNA levels were associated with longer TTP in the pooled sample (log-scale FR = 0.80; 95% CI 0.69-0.94) and specifically in women from Greenland (log-scale FR = 0.72; 95% CI 0.58-0.89). ORs for infertility were also increased in the pooled sample (log-scale OR = 1.53; 95% CI 1.08-2.15) and in women from Greenland (log-scale OR = 1.97; 95% CI 1.22-3.19). However, in a sensitivity analysis of primiparous women these associations could not be replicated. Associations with PFNA were weaker for women from Poland and Ukraine. PFOS, PFOA and PFHxS were not consistently associated with TTP.

CONCLUSIONS

Findings do not provide consistent evidence that environmental exposure to PFAS is impairing female fecundity by delaying time taken to conceive.