In vitro fertilization is a fertility treatment that involves stimulating the ovaries with medications to produce multiple eggs, which are then retrieved around two weeks later. They're either frozen immediately or fertilized to become embryos. It is, without a doubt, the hardest thing I've ever been through. But not for the reasons you might expect.
It wasn't the dozens of injections, blood tests or ultrasounds. It's the emotional side that's the toughest. My wife, family and friends are wonderful, but I didn't want to be that person who only talks about IVF. So, I leaned on ChatGPT to field a lot of my fertility questions (especially because every consultation with my clinic is hundreds of dollars).Â
ChatGPT was extremely helpful during my first IVF cycle. It helped me make sense of all the acronyms and understand the timeline, and I was able to ask for advice along the way. I always made sure to fact-check everything with my doctor.Â
Let me preface again -- ChatGPT is not a doctor. It gets things wrong, and it did in my cycle. It told me I'd need five to eight cycles to get three genetically normal embryos (which is the suggested embryo goal to have a 95% chance of a live birth).
I ended up getting five genetically normal embryos from my second cycle, and I was six months older than my first cycle when I asked ChatGPT. This is significant at my age, where every month matters. So take anything the AI says with a grain of salt.Â
With the recent announcement of ChatGPT Health, we're likely going to see a lot more use cases for AI in the IVF process. ChatGPT has stated you can feed it your medical records (and that the data isn't used to train the model), but I wouldn't trust that. I don't know where my data will end up or what it'll be used for.Â
Sam Altman, CEO of OpenAI, which developed ChatGPT, once said ads in ChatGPT would be "uniquely unsettling," but he has now changed his tune, with ads to roll out this year. Please don't trust big tech with your medical records.
This is a good refresher if you're unsure how much to tell ChatGPT. Treat AI chatbots as public environments to protect yourself.Â
(Disclosure: Ziff Davis, CNET's parent company, in April filed a lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.)
How I used AI in my second IVF cycleÂ
The egg retrievalÂ
I used ChatGPT a little differently for my second IVF cycle. I had a good understanding of all things IVF by now, so this time,Â
I focused more on expectation setting. I switched clinics -- to one of the options ChatGPT actually suggested during my financial research -- where I had easy access to all of my bloodwork and ultrasound data. I fed these into ChatGPT as I progressed through my cycle. Not the downloads, just the raw data.Â
Yes, this was technically personal medical data, but it wasn't sensitive information like my clinic name, ultrasound photos, donor number, etc. It was information like follicle count and sizes, hormone numbers and instructions. This information is unidentifiable, and it doesn't reveal my clinic.Â
By feeding it my follicle count, size progression and hormone numbers after every monitoring appointment, it better understood my growth rate and what to expect at the egg retrieval. My follicles (the sac that contains the egg) ranged from 12mm to 24mm. Follicles above 18mm at the time ovulation is triggered with a medication (for precise egg retrieval timing) tend to produce a mature egg. ChatGPT was able to predict the number of eggs based on the growth between trigger and retrieval, which was a 35-hour time span.Â
AI is great for averages, but remember, it's not always accurate. Based on my data, it predicted 12 to 14 eggs, with 7 to 9 being mature. I ended up with 11 eggs, but 10 were mature. Seven of these fertilized (which was the same as my first cycle, where I ended up with two embryos).Â
In this second cycle, ChatGPT predicted I would get three euploids, which are genetically normal embryos.
A blastocyst is an early development milestone, a rapidly dividing ball of cells. A euploid is a blastocyst that has been tested (via PGT) and is genetically normal. Aneuploid is an abnormal blastocyst.
I ended up with five genetically normal embryos, so I tracked much better than past stats and averages for my age.Â
Bodies are not machines, and when it comes to fertility, expect the unexpected. It's hard to predict, as every month and cycle can be vastly different.Â
I didn't put my embryo report from the genetics company into ChatGPT. I uploaded the results manually like this:
The percentages are the live birth rate based on the embryo day and grading. The number after the day represents the stage of expansion and development of a blastocyst, from 1-6. The first letter is the inner cell mass (cells that become fetus). The second is the trophectoderm (cells that become placenta).
I was able to chat through what this meant, like how these odds are per embryo, not cumulative.
The failure rate is calculated based on my three top embryos that have a 70%, 70% and 65% chance of implantation. ChatGPT calculated it like this: "Failure of each = 30%, 30%, 35%. Chance all fail = 0.30 × 0.30 × 0.35 = 3%. Therefore, the chance at least one works = 97%."
While nothing is guaranteed, knowing I have a 97% shot of one of my top three working, I can comfortably move on to an embryo transfer instead of trying to bank more embryos.
These are the types of questions you can unpack with ChatGPT.Â
It created handy charts like this, too:
I asked ChatGPT to explain grading because I'd heard that if an embryo is euploid, the grading is less relevant.
This was helpful framing for family planning.Â
My wife and I were considering transferring one of the lower-quality grade embryos (a BC grade), which would be a female baby, but ultimately decided to let the embryologist choose our highest-graded embryo (an AA grade).Â
The frozen embryo transfer
Once I started the frozen embryo transfer cycle, I put my data in for my first monitoring appointment:
At every monitoring appointment, the clinic checks E2 (estrogen), progesterone and LH (luteinizing hormone) to time the embryo transfer.Â
The initial test also included three thyroid tests (T3, T4 and TSH). It picked up on something the doctor didn't.
My TSH, which is the thyroid-stimulating hormone, had increased from 2.49 to 2.87, and they like it under 2.5. My doctor put me on a medication called levothyroxine to help reduce it in the lead-up to my embryo transfer.
They also asked me to take baby aspirin, so I asked ChatGPT about that.
Endometrial receptivity means the uterine lining is ready for implantation. There's a "window of implantation" that happens post-ovulation. It's supported by progesterone medications.Â
I like the little summaries it gave.Â
A receptive endometrium means the same as endometrial receptivity.
I kept asking random questions throughout my cycle, like what are the most important things I can do to make an embryo stick, should I continue taking all my vitamins, should I eat beetroot and pomegranate, etc. (Beetroot and pomegranate juice are cited often in IVF circles as good post-transfer foods.) I got a bunch of great tips.
ChatGPT continued giving me advice during the transfer prep, especially because I was doing a modified natural protocol, which meant they were closely monitoring my body's natural ovulation. Timing is critical with this protocol. It helped me unpack and understand each decision alongside my clinical team.Â
However, I did reach a point where I stopped asking ChatGPT questions and logged out of IVF groups. IVF is so information-heavy that it can consume you. It did for me.
Two years into my fertility journey, with multiple healthy embryos (finally), I decided to surrender to the process, over to my body and biology to do the rest of the work.