A Bright Tomorrow for Therapeutic Orphans: The Potential for New Approach Methodologies to Bridge Data Gaps for Pregnant Patients

By Matthew Durthaler | July 24th, 2025

While 90% of pregnant patients take some form of medication, adequate safety data is available for less than 10% of the therapies used within the past several decades. Pregnant patients are generally excluded from clinical trials due to ethical considerations regarding the safety of the mother and fetus. This exclusion is necessary to ensure a safe gestational period in patients; however, this practice has left significant knowledge gaps in our understanding of drug metabolism and disease processes within the pregnant population. So much so that pregnant patients have been dubbed therapeutic orphans—a patient group for whom robust therapeutic data is lacking. Though in vitro, in vivo, and ex vivo models have been used to supplement gaps in data, these models fail to accurately mimic the complex physiological dynamics between the mother and fetus. The shortcomings in these models in conjunction with the general exclusion of human testing suggest that obstetrics may prove to be the ideal frontier on which New Approach Methodologies (NAMs) can unequivocally establish their necessity among clinical research methods.

Pregnancy is a strenuous condition of the body that is accompanied by considerable changes in physiology. This often necessitates additions or alterations to pharmacotherapies.   Unfortunately, safety information is often limited. Novel therapeutic agents may pose unknown risks to both the patient and the fetus. And in vivo testing methods to quantify fetal exposure such as fetal cord blood sampling, are invasive and further risk fetal wellbeing. Due to the risks and ethical considerations, pregnant patients are only included in clinical trials for a narrow set of circumstances that are subject to additional regulation. Such narrow circumstances for research are significantly outpaced by the need for clinical data. The increasing incidence of chronic conditions and the increasing age seen in pregnancy contribute to growing pharmacotherapy utilization within this population. Unfortunately, current models have failed to yield enough data to meet the growing demand.

Due to the insufficient quantity of pregnancy-related data, scientists and clinicians have attempted to bridge some of the limitations. Efficacy and dosing data in pregnant patients is generally extrapolated from clinical trials in non-pregnant humans. Much of the safety data for pregnant individuals is gathered from observation, case-control studies, or post-marketing surveillance data. Unfortunately, these information gathering efforts are of limited value. Extrapolating data from non-pregnant humans fails to account for the physiological changes that occur in pregnancy such as hemodilution, changes in body habitus, increased glomerular filtration, and alterations in metabolism. Further, variations in disease concurrence, degree of fetal development at exposure, and additional medications confound observational data and make it difficult to draw definitive conclusions. Given these limitations, physicians often take a cautious approach to medicating pregnant patients, balancing the known benefits with risks based on available data. This can still be an unsettling experience for both patient and provider because without sufficient research, there remains a degree of uncertainty. As caution often precludes pregnant humans from serving as models of their own physiology, various non-human animal models have been employed. While these models have displayed some value, they are severely limited by interspecies variation and thus serve as subpar predictors of gravid human physiology.

Much of the pregnancy-related experimentation revolves around the placenta—the organ that serves as the barrier between mother and fetus. This organ is highly complex and changes throughout gestation. Thus, models that exhibit similarities in placental anatomy and physiology are necessary for predicting outcomes in humans. In vivo experimentation on pregnant non-human animals is used to account for the placenta’s dynamic nature, but interspecies variation limits its applicability. Humans share their hemomonochorial placental anatomy with non-human primates and guinea pigs; however, pregnant mice are reported to be the most common model. Interspecies variation makes human gestation and rodent gestation broadly incomparable. Rodent pregnancies are short and produce multiple neurologically underdeveloped offspring. In contrast, human pregnancies generally consist of a 40-week gestation period that produces a single offspring. Sheep gestational models are more like humans, as their pregnancy also consists of a single offspring that develops over a multi-month period. Sheep, however, are not effective models since their placental tissue structure differs substantially from humans. Some non-human primates exhibit similarities in both placental anatomy and gestation; however, endocrine or metabolic variation may still present obstacles. Additionally, both ethical and financial costs increase as higher-order non-human animals are employed. Non-human primates that exhibit similar anatomy, gestation, and metabolism are generally cost prohibitive, impeding testing in sufficient numbers to obtain adequate statistical power. Therefore, non-human primates may not provide significantly more information than human clinical trials due to lack of subjects. No single in vivo model can provide adequate information, and despite the availability of multiple different models, the demand for drug safety data has outpaced the supply. Clinical trials in pregnant humans are rare; high throughput, non-human animals face challenges to external validity; and models with minimized interspecies variation are cost prohibitive.

Ex vivo models produce their own set of challenges. While non-human animal placentae can be studied ex vivo, interspecies variation would create the same limitations seen from non-human animal models in vivo. However, certain ex vivo models can control for interspecies differences. A recently delivered human placenta can be studied in a static or dynamic environment. Placental villi explants are an example of one such method. Villi explants consist of a sectioned placenta sample studied in a static environment. These explanted samples preserve the tissue structure and metabolic activity as was displayed in utero immediately prior to delivery. While it displays utility for certain placental transport experiments, the static fluid of villi explants fails to accurately mimic the blood flow within the human placenta. Placental perfusion studies remedy this deficit by connecting a placental lobule to perfusion pumps. These pumps mimic maternal and fetal circulation by driving blood through the placental vasculature, replicating the dynamic physiology of a placenta in utero. This model is generally considered the current gold standard as it may be best at recapitulating the conditions immediately preceding delivery; however, it is limited to this period. In addition to their rapidly waning viability, a severe limitation of ex vivo models is that they generally only mimic the functioning of a placenta as delivered, which is generally the third trimester. Placentae exhibit different levels of permeability between the third trimester and first trimester and since the first trimester is the critical period of fetal development, ex vivo models fail to provide the most critical information about drug effects in early pregnancy.

In vitro placental models ameliorate some of the barriers posed by in vivo and ex vivo experiments and with the addition of NAMs, have the potential to answer unresolved questions. Older in vitro models utilize human-derived specimens such as placenta derived tissue or cell lines studied in a static environment. These models exhibited similar limitations to other in vitro models such as absence of laminar flow or limited dimensionality and complexity. While these previous in vitro models provided a low cost, high throughput method of research, they cannot accurately depict the dynamic and complex environment of the fetal-maternal interface. NAM models such as placenta-on-a-chip have the potential to remedy failures of previous in vitro systems and overcome the limitations of in vivo and ex vivo methodologies. The dynamic flow system and biomimetic tissue layers of placenta on a chip have shown similar results to transport experiments seen in placental perfusion studies. Drug safety studies are only the beginning of what placental NAMs could provide. Current research has displayed recapitulation of glucose transport, pharmacotherapeutic transport, bacterial infection, and even preeclampsia. Soon we may not only have more information on pharmacologic safety data, but also a better understanding of transplacental infections and pregnancy-specific disease processes.

Pregnancy is often regarded as both a joyous and terrifying time for soon-to-be-new parents who are constantly bombarded with sometimes conflicting information and advice. Medical providers and scientists have thus far failed to completely answer all their questions and address many concerns due to lack of scientific studies. While placental NAM technology is still in its infancy, it has the potential to provide confident, evidence-based answers to millions of patients. By accurately replicating the biological relationship between mother and fetus, the placenta-on-a-chip has the capacity to limit the hardships faced by human and non-human mothers alike.

The views expressed do not necessarily reflect the official policy or position of Johns Hopkins University or Johns Hopkins Bloomberg School of Public Health.