We've updated our Privacy Policy to make it clearer how we use your personal data. We use cookies to provide you with a better experience. You can read our Cookie Policy here.


Calculating Gestational Age With Metabolomics

Listen with
Register for free to listen to this article
Thank you. Listen to this article using the player above.

Want to listen to this article for FREE?

Complete the form below to unlock access to ALL audio articles.

Read time: 4 minutes

Sadly, preterm death is one of the leading causes of death among children under five years of age. To develop a solution, it is imperative that scientists can measure the extent of the issue at a population level across the globe. In underdeveloped countries where pregnant women do not have routine access to ultrasound technologies, this can be a challenge.

Researchers from Ottawa Hospital, Canada, supported by the Bill & Melinda Gates Foundation, have developed a mathematical model that may offer an alternative to relying on ultrasound information for calculating gestational age. Described in a paper published in eLife, the model uses a simple blood test to determine a metabolic fingerprint, consisting of unique patterns of specific molecules circulating in the blood, that can help estimate gestational age.

We spoke with Kumanan Wilson, M.D., an internal medicine specialist and senior scientist at the Ottawa Hospital Research Institute, and Lindsay Wilson, an epidemiologist at Ottawa Hospital Research Institute, to learn more about the model and its potential applications.

Molly Campbell (MC): Why is it important to develop a variety of methods for measuring gestational age?

Lindsay Wilson (LW):
Around the world, preterm birth is the leading cause of morbidity and mortality among children under 5. Accurate knowledge of a child’s gestational age can help to direct care for that infant, as it allows clinicians to distinguish between newborns who are preterm and those who are small for gestational age. This is an important distinction as it affects both clinical decision-making and expectations regarding the achievement of developmental milestones. At the population level, knowledge of how common preterm birth is in a particular area allows for effective allocation of resources and permits the evaluation of programs intended to reduce the burden of preterm birth.

Prenatal ultrasound is the gold-standard for measuring gestational age, but in many jurisdictions, particularly in low-income settings, ultrasound technology is not readily available, and issues of recall and documentation make last menstrual period an unreliable measure of gestational age. Gestational age can also be estimated through physical and neurological assessments of the infant postnatally (e.g., Ballard and Dubowitz scores), but these measures are often inaccurate, particularly among preterm infants and those who are small for gestational age, and they are often highly variable between clinicians. Given these limitations, improved methods for estimating gestational age are very important in order to monitor and respond to the local and global challenges associated with preterm birth.

MC: Your research uses “metabolic fingerprints” to estimate gestational age. What type of molecules do you analyze, and how does this information infer gestational age?

Our work uses the analytes collected during routine newborn screening to estimate gestational age. In high-income countries, including Canada, nearly all infants are screened for rare, treatable conditions. This screening is done by measuring levels of a variety of analytes contained in a few drops of blood taken from the infant’s heel. These analytes include amino acids, acyl-carnitines, endocrine markers, and other metabolic and genetic markers. The levels of these analytes can identify potential cases of a variety of diseases, including inborn errors of metabolism, endocrine disorders, and hemoglobinopathies. Additionally, these markers vary considerably by the gestational age of the infant, and our previous research has demonstrated that preterm infants are metabolically distinct from term infants. By combining the analyte data obtained during newborn screening with clinical and demographic variables, our group has successfully demonstrated that these analytes can be used to accurately estimate gestational age to within ±1-2 weeks.

MC: What technologies and analytical methods are most commonly used in your research and in the model?

Kumanan Wilson (KW):
Our research is dependent on the use of newborn screening to measure the levels of analytes in a newborn’s blood. This requires the collection of dried blood spot samples taken from a newborn’s heel or umbilical cord, and the analyte levels are measured using such tools as high-performance liquid chromatography and tandem mass spectrometry. Our statistical analysis consists of multivariable regression modeling that combines analyte data, clinical data, and interaction terms to explore which factors are most strongly associated with gestational age when comparing the model-derived estimates to those obtained through prenatal ultrasound. 

MC: In the press release, you state that you are “exploring the use of artificial intelligence” to make the test even better. Please can you expand on this?

KW: While our models perform quite well currently, there are several limitations that restrict their utility. Currently, our models are based on cohorts of infants where the vast majority are born at term, thereby biasing the model’s performance. As gestational age decreases, as does the accuracy of our model. Additionally, our models currently rely on the inclusion of birthweight as a predictor of gestational age, which limits its utility in infants who are small for gestational age. Our team has begun exploring the use of artificial intelligence and machine learning to improve the accuracy of our models by leveraging these complex analyses capable of handling much larger datasets with many more predictors. Our preliminary results have already demonstrated that the use of these techniques have considerable potential in improving the accuracy of our models.

MC: Your research study focused on applying the model in 1000 pregnant women in Bangladesh. Do you anticipate that the results will differ when the model is applied in different populations of women?

We have previously explored the possibility of developing ethnicity- or region-specific gestational age estimation models. Birthweight, which is a major covariate in all of our gestational age models, is strongly correlated with gestational age and varies considerably by ethnicity. As our original model (which this study sought to validate) was based on a predominantly white cohort of infants, we conducted a retrospective validation study using data from Ontario infants whose mothers were landed or non-landed immigrants. Our results indicated that while tailored algorithms may slightly improve the accuracy of our models, the Ontario-derived model works well among infants from a wide variety of ethnic backgrounds. As such, we do not anticipate any major differences when our models are applied to other infant cohorts during the next phase of our research.

MC: What challenges do you face in implementing this model into modern medicine? How do you anticipate overcoming these challenges?

The approach we have described offers promise for improving post-natal estimation of gestational age. Further work needs to be done to further evaluate generalizability to multiple low-resource settings.  We are currently undertaking this work in partnership with Stanford University. Furthermore, both the cost-effectiveness and ongoing feasibility of this approach needs to be further assessed. Ideally the cord blood approach in combination with point-of-care diagnostics could lead to a test that could yield valid real time results that could guide the care of the newborn.

Kumanan Wilson and Lindsay Wilson were speaking to Molly Campbell, Science Writer, Technology Networks.