Anup Challa, Principal Investigator of the research group Modeling Adverse Drug Reactions in Embryos (MADRE), talks about how using data from electronic health records to simulate drug trials for pregnant women could offer alternatives to delivering babies preterm in the COVID-19 age.
Right now, a lot of the advice for practicing obstetricians and gynecologists is to deliver pregnant women who have COVID-19. You can imagine that this is problematic because many of these women are preterm which means they’re giving birth to preterm babies who require advanced care and oftentimes the effects of preterm births on newborns can last throughout the course of their life. Is there a certain timeframe?
Anup Challa: Often, these decisions are based on a physician’s judgment of risk versus benefit. Certainly, if an obstetrician encounters a first-term patient and feels that delivery at that stage would be more detrimental than potentially continuing with the course of the infection, then that physician would not recommend delivery. But, we are seeing that as women progress further into pregnancy—towards the end of the second trimester and into the third trimester—that those judgments are erring more towards delivery than they are towards any type of therapeutic intervention.
Is this only for women that come to the hospital or if a pregnant woman’s at home and she has sniffles and runny nose? Would it be worth it for her to go to the hospital or is it only for women that come to the hospital with severe cases?
Anup Challa: It goes back to a physician’s judgment call on whether the risks associated with preterm birth warrant delivery, versus allowing a pregnant patient to continue with COVID-19. The concern that we’re hearing more about, is a risk of vertical transmission from a mother to her fetus. If the disease is left untreated, there is worry that the child could be born with COVID-19 as well. Delivering a patient preterm is a significant decision. And it is one that obstetricians are taking with great consideration, because they do not want to reach that conclusion too hastily. It is a judgment call because we do not have any existing standard of care for pregnant women with COVID-19. We do not have a drug we can give them. We do not have clinical trials in which we can enroll them. So, there really is not much recourse for severely infected patients beyond delivering preterm to try to prevent further complications.
How are you guys are working on solving that issue?
Anup Challa: Pregnant women are often excluded from clinical trials for ethical reasons, because an experimental therapeutic might cause harm to their unborn children. Therefore, clinical trials for experimental therapeutics to treat COVID-19 continue to prevent pregnant women from enrolling in them. This furthers a continuing lack of information on the safety and efficacy of drugs that we prescribe to pregnant patients—and, for COVID-19, gives us little knowledge on therapeutic options for severely infected patients and minimal data on preventing the spread of their infections to their children. What we are trying to do is work around this inability to enroll pregnant women in clinical trials by engineering an ethical framework to discover new therapeutic insights through strategic analysis of their existing healthcare data. By analyzing large numbers of electronic health records, we can use high-powered statistics like machine learning to look at outcomes in cases where pregnant women have been exposed to drugs either similar to the experimental therapeutics that are currently in trial or to the therapeutics themselves, since we’re seeing a rise in drug repurposing studies, where drugs that have been previously used to treat other diseases that pregnant patients might have contracted could have application to treating COVID-19. We can then look at the outcomes that manifested in the newborns of pregnant patients who took these drugs, and we can use machine learning to determine whether prescription of these drugs was causal for the birth defects. In this way, we can derive new therapeutic information and simulate a clinical trial by looking at differences in outcomes among pregnant women given a drug and those who weren’t, controlling our models so we can test the association between this drug exposure and the health of their newborn children. We like to call this framework a “target trial.” It makes use of millions of data points of existing primary healthcare information.
How effective are assimilating clinical trials because you do not actually deal with the people but rather, you are dealing with just the records?
Anup Challa: I think this relies on the context in which we are trying to do this. Certainly, in the case of pregnancy—and for expectant woman who have no existing standard of care—any robust therapeutic information could be helpful in considering the risks and benefits of drug exposure during pregnancy. Target trials are not replacements for randomized, controlled trials or other types of clinical trials, but they could powerfully and systematically uncover new knowledge that can be substantiated through other rigorous forms of experimental validation. We have seen that in previous cases where we applied similar statistical models to Vanderbilt’s 2.8 million electronic health records, we found drug development signals that have held up in experimental models. So, we are confident that this method could allow us to derive new insights on drug safety in pregnancy that could facilitate further interrogation downstream. But, in this population of pregnant patients, for whom we do not have good drug safety and efficacy data, our work is an attempt to try to uncover some of the basic information that should be clear—but is not.
So, is it mainly if a woman must be treated with drugs because there are some cases that typically they do not have to be treated with medication?
Anup Challa: Pregnant women are a very vulnerable population for COVID-19. I think it is important to remember that pregnant women are at an increased risk of infection, given the basic biology of how COVID-19 infects a person. The virus enters via a receptor known as ACE2. And, in pregnant women, we know that their lungs contain an excess of ACE2, compared to a non-pregnant person. This means they are likely at a significantly increased risk of contracting COVID-19. Certainly, when you think about therapeutics for COVID-19, there are a broad range of experimental drugs that are currently in trial. And they cover the lifespan of the disease. So, that would include patients who are mildly symptomatic, symptomatic, hospitalized, and patients who are in a severe state. So, when we are looking at pregnant women in the scope of COVID-19 therapeutics, we are not just considering one stage of the disease. We believe that our target trial framework is sufficiently broad to test different experimental therapeutics along the span of infection.
So, if a pregnant woman contracts COVID-19 she is likely to have complications because she is pregnant, so you need treatment for her?
Anup Challa: Yes. Increasingly, the virulence of the disease is appearing higher in the pregnant population than in the non-pregnant population, because the virus can more easily infect a pregnant woman. These patients’ immune systems are in flux during their pregnancy, and as their fetuses grow, they can naturally become short of breath. These factors could contribute to more severe infection. Therefore, given these increased risks for virulence and disease severity, we need to be paying more attention to how we treat pregnant women with COVID-19, especially given the case reports that show that an expectant woman may not be only at risk for herself if she contracts COVID-19, but she could also pass it on to her child.
How can this be applied for other medical conditions?
Anup Challa: Generally, we’re trying to uncover new insights about the safe and effective management of pre-existing conditions in pregnancy that require therapeutic intervention. This is important, because pregnant women may have to stay on drugs that treat chronic conditions they developed before they became pregnant. Physicians need to be able to make informed risk versus benefit decisions on whether or not to continue treating their pregnant patients with drugs, based on how patients may respond if they are taken off their drugs, versus the risks of harm that the drugs could pose to the patients’ developing fetuses. What we’re trying to do is to use our target trial framework to uncover more rigorous and robust drug safety information than is currently available, so, in the absence of pregnant patients’ participation in clinical trials, there are still sufficient patient data for such risk versus benefit decisions to be accurate. In the long run, we hope that physicians will feel confident in being able to use results from our target trial framework in delineating the risk versus benefit tradeoff that they have to make for every pregnant patient who receives a prescription.
Do you have any data so far?
Anup Challa: Yes, I will share an example with you that demonstrates the scalability of the target trial approach. At Vanderbilt, we have electronic health record information for 2.8 million patients, most of which is accompanied by the patients’ genomes. Our colleagues performed a target trial using this information and found that inhibition of a protein called PCSK9 in pregnant women can lead to the development of neural tube defects in their newborns. They looked at electronic health records for pregnant patients that also include their genetics. What they found was that a naturally-occurring loss-of-function mutation within the gene PCSK9 associated significantly to the development of spina bifida in children whose mothers had this mutation. So, in pregnant patients, if we were to inhibit the protein that arises from this gene—a feature of some antihypercholesterolemic drugs, which lower cholesterol—that may lead to birth defects. A large benefit of the target trials we are currently developing is that we are using artificial intelligence to probe a pregnant woman’s electronic health record and extract the medications that were prescribed to her. We don’t have to rely on genomics as a proxy for drug exposure, so we can use the target trial framework to explicitly test casual associations among drugs prescribed to pregnant women and adverse outcome that developed in their children.
What do you think this will mean for pregnant women to know that if they get pregnant that there’s research going on to see what drugs can help them and which drugs they should not take?
Anup Challa: We hope this research will be inspiring for pregnant women. Pregnancy can be a time of confusion—for both patients and their providers. Expectant women are consistently bombarded with conflicting guidance on what they should take and what they should avoid while pregnant. But, often, that information is not derived from robust drug safety data. We hope that our research will allow pregnant patients to feel empowered in knowing that there is ongoing work to improve their standard of care and that they are in capable hands managing their pregnancy and looking out for both their health and the health of their children.
Interview conducted by Ivanhoe Broadcast News.
END OF INTERVIEW
This information is intended for additional research purposes only. It is not to be used as a prescription or advice from Ivanhoe Broadcast News, Inc. or any medical professional interviewed. Ivanhoe Broadcast News, Inc. assumes no responsibility for the depth or accuracy of physician statements. Procedures or medicines apply to different people and medical factors; always consult your physician on medical matters.
If you would like more information, please contact:
Craig Boerner
Vanderbilt University Medical Center
Sign up for a free weekly e-mail on Medical Breakthroughs called First to Know by clicking here
