Industries / Health / Healthcare / Fertility clinics
Success rates and reporting
Counting every treatment and its outcome in national registers, and publishing success rates so that patients, doctors and regulators can compare clinics.
7
Milestones achieved, from 1991 to 2025.
2
To watch for: announced, not yet confirmed outside.
3
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1991
A register of every cycle
The UK's Human Fertilisation and Embryology Act 1990 set up the Human Fertilisation and Embryology Authority (HFEA) to license and inspect clinics, and requires every licensed clinic to send it treatment information. The HFEA has recorded every IVF and donor insemination treatment in the UK since 1 August 1991, about 2 million IVF cycles by its count, and around 100,000 cycles of treatment and storage a year recently.
Credit
UK Parliament; HFEA
1992
US clinics must report
The Society for Assisted Reproductive Technology (SART) began collecting clinic results in the 1980s and published its first report in 1988. The Fertility Clinic Success Rate and Certification Act, enacted on 24 October 1992, requires every US IVF program to report its pregnancy success rates each year through the Centers for Disease Control and Prevention (CDC). CDC published the first national report in 1997 and reported clinic-level data from 1995 through 2022, usually two to three years late.
Credit
Senator Ron Wyden, who sponsored the act; SART; CDC
2001
Europe pools its registers
The European Society of Human Reproduction and Embryology (ESHRE) published its first European IVF-Monitoring (EIM) report in 2001: 203,893 cycles started in 1997 at 482 clinics in 18 countries. It found that 29.6% of IVF deliveries were twins or more, and that triplet rates ranged from 0.4% to 11.9% between countries. The International Committee for Monitoring Assisted Reproductive Technologies (ICMART) now joins registers worldwide; its report on 2019, published in 2025, counted 3,544,683 cycles and 783,073 babies, with one embryo in 55.6% of fresh transfers using the woman's own eggs.
LessonPooled world figures arrive about six years after the treatment, so they show trends, not current practice.
Credit
K. G. Nygren and A. N. Andersen for ESHRE; Silke Dyer, Georgina M. Chambers, G. David Adamson and colleagues for ICMART
2003
Belgium pays for single embryos
From July 2003 Belgium paid the laboratory costs of up to six IVF cycles for women up to age 42, in exchange for limits on how many embryos are transferred; women under 36 in their first cycle get one. At one Belgian clinic, single-embryo transfers rose from 14% to 49% of transfers, the pregnancy rate held at 36% to 37%, and twin pregnancies fell from 19% to 3%.
LessonPaying for cycles in return for fewer embryos per transfer cut twins without cutting pregnancies, at least at this clinic.
Credit
the Belgian government; S. Gordts and colleagues, Leuven Institute for Fertility and Embryology
2017
UK multiple-birth target met
When the HFEA launched its "One at a Time" campaign in 2007, one IVF birth in four was twins or more, 20 times the rate after natural conception. From January 2009 it set a maximum multiple-birth rate that each clinic must not exceed, lowered every year, and let each clinic choose how to meet it. The national rate fell from 24% in 2008 to the 10% target, first met in 2017, and the HFEA says birth rates did not fall.
LessonA cap on the outcome, with the method left to each clinic, cut multiple births across the country.
Credit
HFEA
2017
Add-ons rated by evidence
"Add-ons" are optional extras sold alongside IVF, often at extra cost. An HFEA committee paper of June 2018 says the HFEA first published traffic-light ratings for them in summer 2017; the HFEA's 2023 press release dates its "world-first rating system" to 2019. In 2019 it changed its rating of PGT-A from amber to red. Since October 2023 it uses five colors; its current list rates PGT-A, endometrial receptivity testing and platelet-rich plasma red, time-lapse imaging black (no effect on the chance of a baby), and many others grey (not enough evidence).
Credit
HFEA and its Scientific and Clinical Advances Advisory Committee
2025
US reporting team cut
In early April 2025 layoffs at the Department of Health and Human Services eliminated the CDC's six-person ART surveillance team in Atlanta, NPR reported. The American Society for Reproductive Medicine (ASRM) wrote in 2026 that all its members were laid off and that no replacement federal reporting had been announced; SART still publishes data from its member clinics, which are not all US clinics. CDC's open-data portal has since posted clinic-level tables for 2022 (458 clinics) and 2023 (462 clinics), credited to its Division of Reproductive Health and last updated on 19 November 2026; this search found no national report with them and no statement of who now prepares them.
LessonA legal duty to report depends on a paid team to collect, check and publish the figures.
Credit
NPR's Sarah McCammon (reporting); ASRM
Watch for
Europe counts cycle by cycle
ESHRE says its EuMAR registry records each treatment cycle rather than yearly totals, so it can measure cumulative live-birth rates over several attempts and treatment across borders. A six-month pilot in Estonia, Germany, Portugal and Slovenia ran in 2024 and was validated in 2025, and a two-year second phase is to open it to other countries.
Achieved whenEuMAR publishes cumulative live-birth rates for at least the four pilot countries in a report that national registers or outside researchers can check against their own figures.
Claimed by
ESHRE, 2025 and 2026
Australia's independent clinic accreditation
After several embryo and sperm mix-ups, Australia's health ministers accepted all recommendations of a rapid review in September 2025. Clinics are to be accredited by the Australian Commission on Safety and Quality in Health Care instead of a committee run by an arm of the industry's own society, states are to align registration and reporting rules, and the government-funded YourIVFSuccess website is to add costs and evidence on add-ons. Ministers said the new accreditation requirements "will be in place by January 2027".
Achieved whenclinics are being accredited under the Commission's standards and YourIVFSuccess shows costs and add-on evidence.
Claimed by
Australian health ministers, September 2025
Open
01
One table of world registers
A machine-readable table of the yearly headline figures from every national and regional IVF register that publishes them, such as CDC and SART in the US, the HFEA, ANZARD for Australia and New Zealand, Japan's JSOG, the Latin American registry, ESHRE's EIM and ICMART. Each figure carries its definition: per cycle started, per egg retrieval, per transfer or cumulative; with or without frozen transfers; and what counts as a birth.
The briefCredit · unclaimed02
Advertised rates against registers
A public check of IVF clinic websites in the UK and the US: collect each success rate a clinic advertises, find the matching figure for the same period and patient group in the HFEA's or CDC's published clinic data, and record whether they agree, differ, or cannot be matched because the claim does not say what it measures.
The briefCredit · unclaimed03
One US table since 1995
One tidy, versioned table of every clinic-level figure CDC has published since 1995, with each clinic linked across years through name changes, mergers and closures, each year's definitions recorded, and the code that rebuilds the table from CDC's own files.
The briefCredit · unclaimed
Open 01 · the brief
One table of world registers
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A machine-readable table of the yearly headline figures from every national and regional IVF register that publishes them, such as CDC and SART in the US, the HFEA, ANZARD for Australia and New Zealand, Japan's JSOG, the Latin American registry, ESHRE's EIM and ICMART. Each figure carries its definition: per cycle started, per egg retrieval, per transfer or cumulative; with or without frozen transfers; and what counts as a birth.
Why it is still open
ICMART's world report for 2019 appeared in 2025 as a journal article. A 2017 systematic review hand-built a comparison of seven registers for 2004 to 2013 and found that inconsistencies in how registers present outcomes "made comparison between regions difficult"; it has not been kept up. ESHRE's EuMAR aims to harmonize Europe's data but is still in its second phase.
Built with what exists
AI models that read long PDF reports and pull out tables, such as Claude or Gemini, with a person checking each extracted number against its page; a version-controlled repository; Python for consistency checks. The advance is models that extract tables from long reports and give the page for each value.
First step
Take the latest reports from the HFEA, ANZARD and CDC, extract live births per egg retrieval for one year, and write down how each defines it.
Done looks like
At least 10 registers, every year each has published since 2010, each figure linked to its page in the source report with its definition coded; a second person checks a random 5% of figures and agrees on at least 99%; the table is updated within a month of each new report.
Open 02 · the brief
Advertised rates against registers
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public check of IVF clinic websites in the UK and the US: collect each success rate a clinic advertises, find the matching figure for the same period and patient group in the HFEA's or CDC's published clinic data, and record whether they agree, differ, or cannot be matched because the claim does not say what it measures.
Why it is still open
A 2017 study of 79 UK clinic websites found that 53 reported their success using 51 different outcome measures, usually without sample size or reporting period. SART's advertising committee reviews its US members' websites against its own rules, including whether figures match its clinic reports; in 2019, 44 of 161 clinics reviewed had violations not fixed within two weeks, and 10 of 43 still had violations in 2020. These studies report counts and categories, not a clinic-by-clinic comparison with official figures. OpenIVF, a public directory of Indian clinics on GitHub (last updated September 2026), shows each clinic's reported rate beside its definition, patient group, reporting period and source; it covers India only and does not compare the rates with official figures.
Built with what exists
an AI agent that browses clinic sites and quotes each claim with its wording and date, such as Claude driving Browser Use (see the Note "Deleting 80% to start over"); the HFEA's published clinic figures and CDC's clinic tables on data.cdc.gov; a spreadsheet. The advance is agents that can read hundreds of websites and cite the passage behind each claim.
First step
Take 20 UK clinic websites, extract every advertised rate, and match them by hand to the HFEA's figures.
Done looks like
All HFEA-licensed IVF clinics and at least 300 US clinics checked; every claim archived with the date it was captured; each match or mismatch confirmed by a second person; results published clinic by clinic, with each clinic sent its rows 30 days before publication.
Open 03 · the brief
One US table since 1995
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
One tidy, versioned table of every clinic-level figure CDC has published since 1995, with each clinic linked across years through name changes, mergers and closures, each year's definitions recorded, and the code that rebuilds the table from CDC's own files.
Why it is still open
CDC's clinic data come as separate yearly files whose layout has changed; data.gov lists an older dataset covering 1995 to 2008, and recent years sit in separate tables on data.cdc.gov. Between the 2021 and 2023 success-rate tables, 421 clinic IDs recur and 25 of them carry a different clinic name, such as ID 202, "The New Hope Center for Reproductive Medicine" in 2021 and "CCRM Fertility of Virginia Beach" in 2023, so linking clinics across years needs checking by hand. The team that maintained the series was cut in 2025. A GitHub search in November 2026 found projects that each use a single CDC year, such as a 2026 model fit on the 2021 file, and no combined archive.
Built with what exists
Python and pandas; the Socrata API that serves data.cdc.gov; a coding agent such as GitHub Copilot (see the Note "Effort, not model") to write a parser for each year's layout; a person to confirm each cross-year clinic link.
First step
Download the 2021, 2022 and 2023 success-rate tables from data.cdc.gov and list every clinic ID whose name changed.
Done looks like
Every year from 1995 to the latest posted in one table with a documented column per measure; yearly totals within 0.5% of CDC's national figures; every cross-year clinic link listed with its evidence; code and table public, and a second person rebuilds the table from the code with identical output.