Fertility Technology

In vitro fertilization at work

Machine learning has radically transformed various areas of the healthcare industry, from enhancing drug discovery to improving imaging and diagnostics. One area that has yet to be disrupted is assisted reproductive technology or fertility technology, which is becoming increasingly important as women are having kids later and some are losing eggs sooner. I believe that machine learning can revolutionize this industry by providing estimates of IVF success rates that are far more accurate than traditional methods.

What is IVF?

In vitro fertilization, or IVF, is a process of fertilization that works by using a combination of medicines and surgical procedures to help sperm fertilize an egg, and implant the fertilized egg in a woman’s uterus. IVF is one of many assisted reproductive technologies, but it is by far the most common — a Harris Williams & Co report estimates that IVF accounts for 99% of all assisted reproductive procedures, and that the overall market in the United States is estimated to be worth $1.7 — $2.5 billion.

IVF is especially important for the 12% of women in the United States who experience infertility issues, according to the CDC. IVF is used to treat infertility in the following cases:

• Blocked or damaged fallopian tubes
• Male factor infertility including decreased sperm count or sperm motility
• Women with ovulation disorders, premature ovarian failure, uterine fibroids
• Women who have had their fallopian tubes removed
• Unexplained infertility

What’s the problem?

Despite how common IVF is, it is by no means perfect; the Society for Assisted Reproductive Technology reported an overall 32% success rate in 2015. Furthermore, IVF has many steps, and it takes several months to complete the whole process. It sometimes works on the first try, but many women require more than 1 round or cycle of IVF to get pregnant. Today’s solutions are problematic because couples often don’t have visibility into the number of cycles they need to be successful, and each cycle costs anywhere from $10,000–$15,000 according to Penn Medicine. Due to these issues, more than 50% of women who try IVF for the first time drop out after the first failed cycle, even though their probability of having a baby is high after two or three cycles.

What can technology do?

Machine learning has proven to be an extremely powerful tool for analyzing large data sets and providing predictive estimates, as well as automating low value-added, rote tasks. If adequate volumes of IVF treatments and other data points are gathered, a machine learning algorithm could in theory provide success rates more accurate than traditional age-based estimates.

Univfy is a series A startup that makes that theory a reality. Univfy has raised a total of $15 million to date, and was founded in 2009 by Dr. Mylene Yao, a board certified ob-gyn. Univfy’s solution is a cloud platform that uses machine learning algorithms to better predict the success of IVF, often with an error rate within 5% for most patients. The software is targeted at fertility clinics, which use the software to provide a personalized IVF prognostic report, allowing them to customize financing and refund programs.

Example of Univfy’s PreIVF report

Based on data provided by the clinic (location, population type, previous IVF outcomes, etc.) and access to data collected from over 150,000 IVF treatments, Univfy builds a software-based predictive model that is sent to that clinic via a secure cloud-based system. The PreIVF report is then completed by the doctor, who adds the woman’s medical history. This allows the clinic to set up a refund program that matches the individual patient’s success rate. While refund programs are available on the market, only 10 to 15% of women meet the criteria, whereas women who have access to Univfy’s technology have a much higher chance of qualifying for a refund program (50 to 80%).

Univfy’s proprietary technology was developed at Stanford, and the company is currently working in 12 fertility centers (2.4% of the total 500) that pay a monthly fee to access the software. While there are some competitors in the market, namely Future Family and Fertilome, they target patients themselves and are less effective at estimating multi-cycle success rates.

Ultimately, it’s clear that machine learning has the potential to disrupt the fertility technology market, and companies like Univfy can make it happen.