Trust us, it works!

by Rosie Audino, expert in philosophy and science and health communication

How many times have we heard advice given to pregnant women in the style of the grandma from My Big Fat Greek Wedding, the one who cures everything with a folk remedy and an inevitable "trust me, it works." And it's no coincidence. We thought certain ideas were "grandma stuff," but they actually come from much further back... practically from the Jurassic period. For centuries, pregnancy and childbirth were shrouded in a veil of superstitions and popular traditions that, in fact, contributed to delaying the development of a more systematic and technical study of obstetrics.

In the Middle Ages, pregnancy was entrusted to midwives and healers who acted as "wise women" of the female body. They palpated bellies, listened to heartbeats with their ears, used herbs, oils, and handed-down rituals. It was valuable knowledge, certainly, a practical wisdom built exclusively on daily experience. Incidentally, these women were excluded from official medicine, which consisted exclusively of men! 

And when modern medicine finally tried to address it... things didn't get any better.‍

Thalidomide: The Day Science Understood the Importance of Data (Except for Data Representing Women)

‍Okay, science discovered the importance of data long before the Thalidomide case, but it was crucial for regulating drug approval. Let's find out why. 

In the 1950s, Chemie Grünenthal launched thalidomide after just two months of testing, without adequate animal studies or evaluations of its effects during pregnancy. It was promoted as safe for everyone, including pregnant women. And it was precisely to pregnant women that it was prescribed en masse: it was perfect for first-trimester nausea. Unfortunately, the first trimester is when the fetus develops everything: limbs, organs, and the nervous system. The result was one of the greatest disasters in medical history. Thousands of children were born with severe malformations; the most credible estimates speak of 10,000 - 12,000 affected newborns and an enormous number of spontaneous abortions never formally attributed to the drug. Only in 1961 did two doctors, Widukind Lenz and William McBride, demonstrate the causal link.

After the thalidomide case, the FDA realized that much stricter rules were needed for drug approval. In the 1960s, it introduced for the first time the requirement to demonstrate safety and efficacy through controlled studies, adverse effect monitoring, and informed consent. This was an absolute revolution for the era. However, this turning point also gave rise to a major paradox. In 1977, to avoid any risk of pregnancy during a trial, the FDA recommended excluding all women of childbearing age from the early phases of clinical trials. The result was that for almost twenty years, drugs were tested almost exclusively on men, and then prescribed indiscriminately to the entire population. In essence, in an attempt to protect women, research ended up erasing them. 

Only in the 1990s did the FDA reverse its stance, finally making the inclusion of women in studies mandatory. However, it was observed that even when women were included, many studies did not differentiate data by sex: an FDA analysis showed that one-third of the documents did not report disaggregated data, and 40% did not even indicate the sex of the participants. 

It is precisely because of this exclusion that an entire generation of drugs was developed without adequate data on the female body, and these drugs continue to be prescribed. Yet today we know that the menstrual cycle alters the effectiveness of antipsychotics, antihistamines, and antibiotics. We also know that some antidepressants work differently depending on the phase of the cycle, or that certain drugs can alter heart rhythm, posing greater risks in the first two weeks of the cycle. These are real, even deadly, risks.

Less Folk Beliefs, More Scientific Data

For a long time, many obstetric practices were adopted simply because "that's how it's always been done." The most striking example is episiotomy: for decades, it was an almost automatic gesture during childbirth, routinely performed with the conviction, never truly proven, that it prevented lacerations and protected the pelvic floor. Only when studies finally measured the results did it become clear that this wasn't true: routine episiotomy increased pain, complications, and recovery times. And it wasn't the only practice that was more traditional than scientific. Even the famous "birthing bed," where women lie on their backs with legs in the air—a position uncomfortable even for breathing—became the standard not because it was the best position for childbirth, but because it was the most comfortable... for the doctor, not for the woman. Squatting, lateral, or vertical postures (all physically more effective) were ignored for centuries. Or the prohibition of drinking or eating during labor, adopted for generations without real evidence, and only revised when studies showed that, for most women, there is no risk. But for decades, no one had truly measured anything; they relied on tradition. And the fact that all this data on pregnancy and childbirth was not collected is a huge problem, because it has left deep gaps in knowledge and, consequently, in the care of many obstetric conditions.

Did you know that…

According to data from the World Health Organization, 830 women die each year due to complications during pregnancy and childbirth  https://www.who.int/en/news-room/fact-sheets/detail/maternal-mortality 

From popular beliefs to unbiased technologies

Now let's fast forward: from the Jurassic to the future. In the future of reproductive and sexual health, there is artificial intelligence, which can become a powerful ally, but only if it's trained with truly representative datasets.

What is a representative dataset? A dataset is "representative" when it contains data that truly reflects the real population: this means including women, pregnant women, people of different ages, origins, and clinical conditions. If the data lacks information about the female body, as has happened for decades in medical research, AI will never truly learn to treat women, because it unintentionally discriminates against them.

"Training" a model means teaching it to recognize hidden patterns in data. In practice, it is shown thousands of examples until it learns on its own what is normal and what indicates a risk. To learn more, I refer you to the article AI wanted to be neutral... but then it discovered our biases

An algorithm that detects preterm birth before doctors do

A recent study recorded the electrical activity of the uterus in hundreds of pregnant women. AI learned to read those signals and predict which pregnancies were at risk of preterm birth weeks before a doctor could detect it. Early detection of preterm birth saves lives: it reduces respiratory complications, neurological damage, infections, and maternal and neonatal mortality. This is only possible because the AI was trained on real data from pregnant women, not on adapted male models. (Xu Y, Zu Y, Zhang Y, Liang Z, Xu X, Yan J. A Multi-Algorithm Machine Learning Model for Predicting the Risk of Preterm Birth in Patients with Early-Onset Preeclampsia. Int J Gen Med. 2025 Aug 4;18:4195-4207. doi: 10.2147/IJGM.S521763. PMID: 40786956; PMCID: PMC12333876)

The most important aspect is that the AI used was trained with real data from pregnant women. Not data derived from studies on men, nor generic information adapted to the female body, but knowledge collected directly from women and their pregnancies. 

It's a concrete step towards reducing the gender data gap. 

And Geen? 

Geen positions itself precisely within this trend: it develops an algorithm trained with representative data, including women of all ages, non-binary people, trans people, and diverse ethnicities. Because only this way can tools be created that truly work for everyone. 

‍