Can data analytics help improve the diagnosis and treatment of diabetes in pregnancy?

Dr Lucy Mackillop is a consultant obstetric physician at Oxford University Hospitals NHS Foundation Trust; Honorary Senior Clinical Lecturer, Nuffield Department of Women’s and Reproductive Health, University of Oxford; and Chief Medical Officer at Sensyne Health. Here, she discusses how to improve the diagnosis and treatment of diabetes in pregnancy.

Diabetes in pregnancy is characterised by high blood glucose levels (hyperglycaemia) which can develop during pregnancy (Gestational Diabetes or GDM) or may already be present prior to pregnancy in the form of pre-existing type 1 or type 2 diabetes. In both cases, if not managed effectively, the condition can lead to serious maternal and neonatal complications during pregnancy and after birth. These can include pre-term birth, birth trauma, an increased risk of stillbirth, as well as pre-eclampsia and an increased likelihood of induced labour or a caesarean section.

Demographic and lifestyle changes worldwide have caused an increase in the prevalence of diabetes in pregnancy, with 20% of pregnant women in the UK affected with hyperglycaemia in 2019. Earlier detection and targeted, personalised interventions are therefore increasingly important and support the UK government’s ambition of halving the rates of stillbirths and neonatal and maternal mortality by 2025.

Irrespective of the continuous global increase in diabetes in pregnancy, monitoring the condition often remains an inefficient and time-consuming process for women and their healthcare teams. The risk of complications can be reduced if hyperglycaemia is detected early and managed effectively through regular recording of blood glucose levels, change of diet, regular routine exercise, and pharmacological treatment only if needed. In fact, around 70% of women who develop gestational diabetes can control their blood glucose just by implementing these lifestyle measures.

Thankfully, today we have access to technology such as AI, machine learning and remote patient monitoring that can help support women and clinicians, better manage this condition.

Lack of standardised screening for GDM

Traditionally, GDM is diagnosed by challenging the body to a high dose of glucose. The screening test, called an oral glucose tolerance test (OGTT), is usually done when the woman is between 24 and 28 weeks pregnant and can be a time consuming and unpleasant process. It involves a blood test in the morning to measure the fasting blood glucose level, after the woman has been asked to fast overnight. She is then given a glucose drink and the blood test is repeated two hours later. If her glucose levels are not controlled below a certain level, the woman is diagnosed with gestational diabetes.

The lack of international standardisation of the gestational diabetes screening process, causes some difficulties. The quantity of glucose given as well as the time at which the woman has a blood test and the threshold of what is considered abnormal differs in different areas of the country and internationally meaning that there is no universally agreed way of diagnosing GDM and this can cause confusion among both patients and clinicians.

GDM can be thought of as a syndrome due to the variety of factors that can impact how a woman manages her glucose levels – from her body fat percentage, her metabolic rate, to her age and genetics. While the traditional approach to diagnosing the disease has involved giving patients the same test and then expecting everyone to behave in the same way, today we have the opportunity to use technologies such as machine learning and data analytics to provide more personalised method of diagnostics and care.

How can we improve GDM diagnosis and treatment?

Using a unique anonymised real-world database of pregnant women with or at risk of diabetes in pregnancy, we can help explore metrics that may better predict individual risk of birth outcomes related to hyperglycaemia and in doing so, challenge the use of OGTTs as the gold standard.

As GDM is a syndrome (a group of symptoms), so different phenotypes or subpopulations exist. Characterising these phenotypes may lead to more individualised care. For example, we need to be able to better understand what the optimum timing is for birth for a woman with GDM. This is likely to vary for different sub populations of women with GDM. By creating anonymised real-world datasets generated by regular monitoring of blood glucose levels using remote monitoring technology, and applying advanced data analytic techniques, we can identify the metrics and patterns that are associated with different outcomes and therefore inform clinicians of individual risk prediction based on routinely collected patient data.

The occurrence and type of complication present in a database allow us to further investigate causal links between specific factors and maternal complications. This type of data analysis empowers clinicians to discover new interventions and treatments that are aimed at reducing these complications in the future.

Furthermore, clinicians may be able to identify, based on data collected during pregnancy, which mothers and their babies are particularly vulnerable to longer-term health conditions such as type 2 diabetes or cardiovascular disease later in life. This can lead to the development of postnatal interventions for the highest risk groups and ultimately a healthier life for everyone.

Technologies such as remote patient monitoring and data analytics have the potential to revolutionise the way clinicians manage women with GDM, allowing them to provide increased personalised treatment and care based on patterns within a woman’s individual data. We need to capitalise on the benefits these technologies can bring to augment patient care and lead to better patient outcomes.

Staff Writer