SART/CDC-style statistical model of live-birth probability across the IVF treatment funnel
Modern IVF success calculators (SART's Patient Predictor, the CDC national ART surveillance model, and the UK HFEA calculator) are built from millions of registered treatment cycles. They combine a handful of powerful predictors — maternal age above all — into a personalized probability estimate for live birth per cycle.
Female age is by far the most powerful predictor in every published IVF outcome model. It captures two biological processes simultaneously: the size of the remaining ovarian follicle pool (quantity) and, more importantly, the rate of meiotic chromosome segregation errors in oocytes (quality).
Women are born with their full lifetime supply of oocytes — roughly 1–2 million at birth, declining to ~300,000 at puberty and continuing to decline (largely independent of pregnancy or hormonal contraception) until menopause. What changes most sharply with age is not just the shrinking pool, but the fidelity of meiosis I chromosome segregation in the remaining oocytes — the direct cause of age-related aneuploidy.
Because age affects egg quality and quantity together, essentially every downstream stage of the IVF funnel — fertilization, blastocyst formation, euploidy, implantation — is filtered through this one variable.
National registry data (SART/CDC-style) show live birth rate per IVF cycle (own eggs) falling from roughly 50% under age 35, to about 38% at 35–37, 25% at 38–40, 12% at 41–42, and under 5% after age 42.
Anti-Müllerian Hormone (AMH), secreted by granulosa cells of small growing follicles, is the most reliable blood marker of the remaining ovarian follicle pool. Antral Follicle Count (AFC), measured by transvaginal ultrasound, is its imaging counterpart.
• AMH > 3.0 ng/mL: high reserve, typically strong stimulation response, some risk of ovarian hyperstimulation • AMH 1.0–3.0 ng/mL: normal reserve for reproductive age • AMH < 1.0 ng/mL: diminished ovarian reserve (DOR) — fewer oocytes expected per retrieval, though egg quality (age-driven) is a separate axis entirely
Critically, AMH predicts how many eggs a stimulation cycle will yield — it does NOT reliably predict egg quality or live-birth probability on its own. A 42-year-old with excellent AMH still faces the aneuploidy rates typical of her age; a 32-year-old with low AMH may need a more aggressive stimulation protocol but her eggs remain chromosomally young.
Beyond age and reserve, prediction models incorporate several secondary factors:
• Underlying diagnosis: tubal factor and male factor infertility carry near-average prognosis once IVF/ICSI bypasses the mechanical barrier; endometriosis (especially stage III–IV) modestly reduces oocyte yield and implantation; unexplained infertility carries a broadly average prognosis • Prior IVF attempts: a history of prior failed cycles at the same age modestly lowers per-cycle prognosis (selection effect — the remaining "easier" cases already succeeded), while it also allows response-based protocol tuning that can improve subsequent yield • Body Mass Index (BMI): obesity (BMI ≥ 30) is independently associated with lower oocyte yield, lower implantation rates, and higher miscarriage risk; being underweight (BMI < 18.5) is linked to ovulatory dysfunction. Most national calculators apply a modest downward adjustment outside the BMI 18.5–29.9 range
These secondary factors typically shift predicted probability by a few percentage points — meaningful, but dwarfed by the effect of maternal age.
Controlled ovarian stimulation (COS) uses exogenous gonadotropins (FSH ± LH) to rescue a cohort of antral follicles from the natural monthly cycle of atresia, growing many follicles to maturity simultaneously instead of the single dominant follicle of a natural cycle.
Each menstrual cycle, a cohort of antral follicles (2–10 mm) begins growing, but natural selection normally allows only one to reach ovulation while the rest undergo atresia. Gonadotropin stimulation suppresses this selection by supplying FSH in excess, rescuing many follicles simultaneously.
Follicles are monitored by serial transvaginal ultrasound (measuring diameter) and serum estradiol. When a cohort of follicles reaches ~17–20 mm, a trigger injection (hCG or GnRH agonist) induces final oocyte maturation, and oocytes are retrieved transvaginally under ultrasound guidance roughly 36 hours later.
Oocyte yield is governed primarily by the size of the recruitable follicle cohort — which is why AMH and AFC, both markers of that cohort, predict oocyte number so well. A woman with AMH of 4.0 ng/mL might yield 15–20 oocytes; one with AMH of 0.5 ng/mL might yield only 2–4, even with maximal stimulation dosing.
Oocyte quantity (driven by AMH/AFC) and oocyte quality (driven by age) are largely independent axes. A high oocyte count in an older patient does not restore the chromosomal fidelity that only younger age provides — it simply gives the euploidy filter more raw material to work with.
Not every retrieved oocyte is ready for fertilization. Oocytes are staged by nuclear maturity:
• MII (metaphase II): fully mature, has completed the first meiotic division and extruded the first polar body — the only stage competent for fertilization • MI (metaphase I): immature, still completing meiosis I — occasionally matures in vitro within hours • GV (germinal vesicle): immature, nucleus intact — rarely usable
Across a typical stimulated cycle, roughly 80–85% of retrieved oocytes are MII-mature, largely independent of age within the reproductive range, though very high or poorly-timed trigger dosing can reduce maturity rates.
Stimulation protocols are individualized using AMH/AFC to predict response and avoid two failure modes:
• Poor response (< 4 oocytes): often seen in diminished ovarian reserve; may require protocol adjustment, dose escalation, or counseling about donor oocytes at very low AMH combined with advanced age • Ovarian hyperstimulation syndrome (OHSS): excessive response (> 20 oocytes) in high-AMH patients (e.g., PCOS) raises risk of a potentially serious fluid-shift complication; mitigated with GnRH-antagonist protocols, agonist triggers, and elective freeze-all strategies
The clinical sweet spot for most prognostic models is 10–15 oocytes retrieved, which maximizes the expected number of usable blastocysts while minimizing OHSS risk.
From the moment of retrieval, each subsequent stage of embryo development filters out a fraction of the cohort. Understanding this funnel — and where age-related attrition concentrates — is essential to interpreting a per-cycle probability estimate.
Intracytoplasmic Sperm Injection (ICSI) — direct injection of a single sperm into the oocyte cytoplasm — is now used in the large majority of IVF cycles worldwide, including for non-male-factor infertility, because it allows precise confirmation of fertilization and works reliably regardless of sperm parameters.
Fertilization is confirmed roughly 16–18 hours later by the appearance of two pronuclei (2PN) — one from the oocyte, one from the sperm — under the microscope. A 2PN zygote is a normally-fertilized embryo ready to begin cleavage division.
Fertilization rates run approximately 70–80% of mature (MII) oocytes with ICSI, and are relatively stable across the reproductive age range — the fertilization step itself is not strongly age-dependent; the downstream stages are where age exerts its dominant effect.
Zygotes divide roughly every 12–24 hours: 2 cells (day 1–2), 4–8 cells (day 3), compaction into a morula (day 4), and cavitation into a blastocyst with a fluid-filled cavity, an inner cell mass (future fetus), and trophectoderm (future placenta) by day 5–6.
Extended culture to the blastocyst stage (rather than transferring at day 3) serves as a natural quality filter: embryos with major developmental competence problems — many of them chromosomally abnormal — tend to arrest before reaching the blastocyst stage. This is one reason blastocyst-stage transfer generally outperforms day-3 cleavage-stage transfer in implantation rate.
Blastocyst formation rates decline with maternal age, from roughly 55–58% of fertilized zygotes under age 35 to around 33–38% after age 42 — a direct downstream consequence of the same meiotic errors that drive age-related aneuploidy.
Each stage of the funnel compounds: starting from 10 mature oocytes at age 38–40, typical yields might be ~7–8 fertilized, ~3–4 blastocysts, and only ~1–2 euploid embryos — illustrating why oocyte number alone is a poor proxy for cycle success after the mid-to-late 30s.
Blastocysts are graded (commonly the Gardner scale) on three axes: expansion stage (1–6, how large/hatched the cavity is), inner cell mass grade (A–C, future fetal tissue quality), and trophectoderm grade (A–C, future placental tissue quality) — e.g. a "4AA" blastocyst denotes an expanded, top-grade blastocyst on both cell lineages.
Morphological grading correlates modestly with implantation potential but is a substantially weaker predictor of chromosomal normalcy than PGT-A testing — a top-morphology-grade blastocyst can still be aneuploid, particularly in older patients, which is why genetic testing (Stage 4) adds independent prognostic value beyond visual grading alone.
Preimplantation Genetic Testing for Aneuploidy (PGT-A) biopsies a handful of trophectoderm cells from each blastocyst and screens the full chromosome complement. The resulting euploidy rate — the fraction of blastocysts with a normal 46-chromosome complement — is the single clearest biological explanation for the age-related decline in IVF success.
Aneuploidy — an abnormal number of chromosomes — in human eggs originates almost entirely during meiosis I, which begins before a woman is even born (oocytes arrest in prophase I during fetal life) and does not complete until decades later, at ovulation. This decades-long arrest is unique to oocytes among human cells.
The cohesin protein complexes that hold sister chromatids and homologous chromosomes together during this arrest degrade progressively over time without being replenished. By the late 30s and 40s, weakened cohesion leads to premature separation of chromatids and non-disjunction during meiosis I, producing eggs with an extra or missing chromosome.
This single mechanism — "cohesin fatigue" over decades of meiotic arrest — is now considered the leading biological explanation for why aneuploidy rates climb steeply after the mid-30s, largely independent of any lifestyle factor.
PGT-A data across fertility centers consistently show blastocyst euploidy rates of roughly 65–70% under age 35, dropping to about 50–55% at 35–37, 35–42% at 38–40, 20–27% at 41–42, and 10–17% after age 42.
A euploid PGT-A result substantially increases the probability that a transferred blastocyst implants and results in a live birth — and, importantly, transferring one euploid blastocyst at a time yields a live-birth rate per transfer that is relatively consistent across maternal age groups (roughly 60–65%), because the test has already filtered out the dominant age-related failure mode.
However, PGT-A does not guarantee a healthy live birth (a small residual risk of mosaicism, testing/biopsy error, or non-chromosomal causes of implantation failure remains), and an aneuploid result does not always mean the embryo would have failed — some mosaic embryos (a mixture of normal and abnormal cell lines) have resulted in healthy live births when transferred after informed counseling, an active area of ongoing research.
The main clinical value of PGT-A is prioritization: in a cohort of multiple blastocysts, testing identifies which one(s) to transfer first, reducing the number of failed transfers and the time to live birth or informed cycle discontinuation — this benefit becomes proportionally larger as maternal age (and aneuploidy rate) increases.
Multiplying the funnel forward makes the age effect concrete. Starting from the same 10 mature oocytes:
At age 32: ~7.6 fertilized → ~4.2 blastocysts → ~2.7 euploid embryos At age 40: ~7.6 fertilized → ~3.2 blastocysts → ~1.1 euploid embryos At age 43: ~7.6 fertilized → ~2.6 blastocysts → ~0.4 euploid embryos
The oocyte and fertilization stages barely move with age — it is the blastocyst and euploidy filters that do almost all of the age-related work, which is exactly why "egg quality" (not "egg quantity") is the dominant clinical concept in counseling patients over 38.
The final output of a prediction model is rarely a single number — patients are counseled on both the per-cycle probability and the cumulative probability across a realistic treatment course (commonly 2–3 stimulation cycles), which is often the more clinically meaningful figure.
A single completed IVF cycle (stimulation → retrieval → fertilization → transfer, including any resulting frozen embryo transfers from that retrieval) carries the age-stratified live-birth probabilities shown throughout this model — roughly 50% under 35 down to under 5% after 42, using national registry-style averages.
Because many patients undergo more than one stimulation cycle, cumulative live-birth probability — the chance of at least one live birth across a defined number of complete cycles — is calculated (in simplified form) as:
Cumulative(n) = 1 − (1 − p)ⁿ
where p is the per-cycle probability and n is the number of cycles. This simplified formula assumes independence and a constant per-cycle probability, which is an approximation — in practice a patient's prognosis, and the clinic's selection of which patients continue treatment, can shift between cycles. Registry-based cumulative outcome studies nonetheless consistently show cumulative rates substantially exceeding any single cycle's probability, which is central to counseling patients on realistic treatment horizons.
Illustrative example: a 36-year-old with a ~40% per-cycle live-birth probability has a modeled cumulative probability of roughly 1 − (0.6)³ ≈ 78% after three complete cycles — nearly double the single-cycle figure.
Tools such as the SART Patient Predictor, the CDC's national ART success-rate reports, and the UK HFEA calculator are built from hundreds of thousands of registered treatment cycles reported annually by fertility clinics under regulatory mandate. Statistical models (commonly logistic regression or machine-learning variants trained on registry data) estimate live-birth probability as a function of age, diagnosis, prior treatment history, and clinic-level factors, then are validated against held-out cycles to check calibration.
These models report population-level probabilities: an individual patient's true probability may differ based on factors not captured by the model (specific genetic conditions, sperm factors, uterine factors, undiagnosed conditions). They are decision-support tools for counseling, not individual guarantees — professional guidelines emphasize presenting a probability range alongside the point estimate.
No model captures every variable that affects outcome. Recognized limitations include:
• Registry lag: national data reflects outcomes from cycles performed 1–3 years earlier, before the most recent laboratory or protocol advances are reflected • Clinic heterogeneity: individual clinic success rates vary by patient case-mix, laboratory quality, and reporting practices — national averages smooth over this variation • Unmeasured factors: sperm DNA fragmentation, uterine receptivity, immunological factors, and genetic conditions in either partner are incompletely captured by standard model inputs • Selection and survivorship effects: patients who return for additional cycles after a failure are not a random subset of all patients, which can bias simple cumulative-rate calculations upward or downward depending on the population studied
Despite these caveats, age-and-diagnosis-stratified prediction models remain the best available tool for setting realistic, individualized expectations before starting treatment.
| Product | Indication | Trial Design | Key Result |
|---|---|---|---|
| Under 35 | ~50% per cycle | Euploidy ~65–70%; strongest prognosis group | ~85–90% cumulative over 3 cycles |
| 35–37 | ~38% per cycle | Euploidy ~48–55%; modest reserve decline begins | ~75% cumulative over 3 cycles |
| 38–40 | ~25% per cycle | Euploidy ~35–42%; blastocyst rate declining | ~58% cumulative over 3 cycles |
| 41–42 | ~12% per cycle | Euploidy ~20–27%; oocyte yield also declining | ~35–40% cumulative over 3 cycles |
| Over 42 | ~4% per cycle | Euploidy ~10–17%; donor oocytes often discussed | ~15% cumulative over 3 cycles |