Pregnancy-related pelvic girdle pain (PPGP) is a common complication of pregnancy affecting between 23 and 65 per cent of pregnant women worldwide. 

Distinct from low back pain, it affects the pelvic joints—in particular, the pubic symphysis and sacroiliac joints. 

Pregnant people who experience PPGP report challenges with walking, working, sleeping, mood and more. 

In 75 per cent of women who experience PPGP during pregnancy, symptoms generally resolve after birth; however, some women experience pain for up to two years following delivery. 

Women who have developed PPGP during pregnancy will usually experience it again in subsequent pregnancies. 

Unlike other pregnancy complications such as gestational diabetes and pre-eclampsia, however, there are no screening programs for identifying pregnant women at risk of PPGP and many people only hear about it after they develop symptoms. 

Current treatments include simple pharmacological pain relief, physiotherapy and exercise programs, and self-management strategies to reduce the impact of the condition. 

More severe presentations can require the use of pelvic support belts, crutches for walking or occasionally bed rest. 

‘When you look at the research and literature for PPGP, the quality of evidence is poor—we don’t know exactly what risk factors are important. 

‘In terms of treatment, there are no interventions that we could say are totally effective,’ says Atefe Ashrafi, a physiotherapy researcher at Western Sydney University who is looking at the risk factors with the aim of developing a tool to screen pregnant women before they develop the condition. 

Previous research by one of Atefe’s co-supervisors, Dragana Ceprnja APAM, looked at the prevalence and risk factors of PPGP in Australian women. 

The research revealed that in a cohort of 780 pregnant Australian women, 44 per cent developed PPGP. 

The odds of developing PPGP increased with each week of gestation (Ceprnja et al 2021). 

In qualitative studies, both the women with PPGP and the healthcare professionals looking after them reported a lack of information about the condition and the need for more education and support for pregnant women who develop it (Ceprnja et al 2022, Ceprnja et al 2023). 

Atefe’s research builds on Dragana’s findings, using machine learning (a form of artificial intelligence) to examine the dataset and pinpoint the risk factors for PPGP. 

While the original analysis of risk factors used traditional statistical methods including logistic regression to examine and identify predictive risk factors, the machine learning approach allows for a sophisticated analysis and identification of patterns in the data (Ashrafi et al 2025). 

The results were published recently in Musculoskeletal Science and Practice. 

PhD student Atefe Ashrafi is using machine learning to analyse research data about pregnancy-related pelvic pain.

‘Machine learning models can give us a better understanding of the interactions between factors,’ Atefe says. 

‘For example, having a higher BMI during pregnancy can influence the amount of physical activity a pregnant woman does as well as engagement with other activities—it’s multi-factorial. 

‘Considering the relationship between the variables might be a better way of looking at the information.’ 

In addition, using machine learning allows the researchers to consider many models rather than just the most likely or traditional models. 

Five machine learning algorithms were compared and the results were measured against the original study, which employed conventional linear and logistic regression models. 

The study showed that two machine learning models (one a form of logistic regression and the other a decision tree-based model) in particular had high levels of predictive accuracy for PPGP and performed better than the previously used conventional models. 

Using these models, the most significant predictive factor was identified as a history of low back pain or pelvic girdle pain in the past, including during previous pregnancies. 

‘If a pregnant woman has had pelvic pain and low back pain in her previous pregnancies, she is more likely to develop that pain in her subsequent pregnancies,’ Atefe says. 

Another important factor is the person’s family history of PPGP, which she says suggests that there may be a genetic component involved as well. 

Other key predictive factors include gestational age and long periods of standing during the day. 

Non-significant predictive factors include maternal age and having a twin pregnancy. 

‘There are two different types of risk factors—some are modifiable and some are not. 

‘I hope that in the future we can give pregnant women personalised advice. 

‘If someone has had pain in a previous pregnancy, we can’t do anything about the past pain but we can say that it is something to take care of during this pregnancy,’ she says. 

The ultimate goal of the project is to develop a tool that can be used to stratify the risk of developing PPGP during pregnancy, preferably before people start experiencing pain from the condition. 

Atefe is using a Delphi study to refine the approach and identify the best questions to ask in a screening tool. 

She expects a prototype to be ready for testing and validation towards the end of this year. 

‘We’d like to do some pilot studies and longitudinal studies to see if the women with risk factors that our screening tool identified go on to develop PPGP,’ Atefe says. 

Quick links: 

• references 

Course of interest: 

It’s time to talk about the pelvis: integrating pain education in the MX of pelvic pain