Does correcting for bias caused by unequal survival in the treatment arms of a randomised controlled trial matter?

Anwar T Merchant and Bryn E Davis

Merchant Davis

The Obstetrics and Periodontal Therapy (OPT) Study was an NIH-funded randomised controlled trial designed to evaluate whether periodontal treatment in pregnant women had any effect on preterm birth; its findings were published in 2006. The investigators randomly assigned about 800 women who had been pregnant for less than 16 weeks, and had periodontal disease, to one of two groups. One group received periodontal treatment during pregnancy, whereas the other group received treatment after pregnancy.

Although the study found that treatment controlled periodontal infection and reduced the microorganism load, there was no difference in preterm birth rates between the two groups. The investigators concluded that treating periodontal disease during pregnancy did not affect the risk of preterm birth. However, they also found that there were more stillbirths in the group that received treatment after pregnancy, suggesting that periodontal treatment may improve survival of fetuses. The potential bias resulting from the intervention affecting both the outcome (in this case, preterm birth) and survival (in this case, stillbirth) was acknowledged as a limitation.

Since then, newer epidemiological methods have been developed to quantify the bias that can occur in a randomised controlled trial when the intervention affects both the outcome and survival. As the OPT investigators had made data from their study available in the public domain, we were able to reanalyse their data, applying these newly developed methods, in our study published recently in the International Journal of Epidemiology.

To apply these new methods, the study question is conceptualised as a hypothetical. In this case, suppose we compared preterm birth risk among women who had live births in both treatment groups in the OPT study if – hypothetically – both groups received treatment during pregnancy (contrary to what actually happened) (Figure 1). The difference in preterm birth risk between the two groups in the hypothetical comparison would then be the degree of bias that was caused by stillbirths, which could be used to correct for bias in the clinical trial estimates. Because these methods test the hypothetical, they are called ‘counterfactual’ or ‘potential outcomes’ methods.

Merchant_Fig1
Figure 1. (A, left) What actually happens in an RCT when treatment benefits fetal survival. (B, right) Mothers who have live births in the RCT. If, hypothetically, all these mothers receive periodontal treatment, more preterm births would be expected in the treatment group because it has more high-risk mothers. The difference in preterm birth risk between the treatment and control groups is the excess risk caused by pregnancy losses in an RCT when treatment benefits fetal survival. (Credit: Bryn Davis)

In our study, we reanalysed the OPT data using counterfactual methods and found that if women with periodontal disease were treated for the disease before 20 weeks of pregnancy, they may have a lower risk of preterm birth.

Merchant_Fig2.JPG
Figure 2. Hypothesised mechanism through which periodontal infection can cause adverse birth outcomes. Periodontal microorganisms proliferating in the mouth enter the bloodstream and, via the umbilical arteries, infect the placenta. (Credit: Bryn Davis)

Periodontal infection is controlled by treatment that is safe and well-tolerated. However, many people with early periodontal disease may not know that they have it (Figure 2).  Our findings suggest that encouraging women to visit a dentist, either when they are planning to get pregnant or in their fifth or sixth month of pregnancy, may help prevent adverse birth outcomes.

Read more:

Merchant AT, Sutherland MW, Liu J, et al. Periodontal treatment among mothers with mild to moderate periodontal disease and preterm birth: reanalysis of OPT trial data accounting for selective survival. International Journal of Epidemiology, dyy089, https://doi.org/10.1093/ije/dyy089.


Anwar T Merchant is Professor and Epidemiology Division Director in the Department of Epidemiology and Biostatistics, Arnold School of Public Health, University of South Carolina, USA.

Bryn E Davis is an artist, with a Bachelor of Science in Microbiology and a Masters in Epidemiology, working for the Disability Research and Dissemination Center at the University of South Carolina.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

w

Connecting to %s