The declining efficiency of IVF in the USA

Abstract

STUDY QUESTION

Has the efficiency of IVF cycles in the USA declined between 2012 and 2021?

SUMMARY ANSWER

National U.S. data show a continuous linear decline in IVF cycle efficiency between 2012 and 2021.

WHAT IS KNOWN ALREADY

Previous studies reported that live birth rates after IVF increased until 2010 but subsequently declined, potentially due to the introduction of IVF ‘add-ons’ and changes in clinical practice.

STUDY DESIGN, SIZE, DURATION

Retrospective cohort study, using publicly available national U.S. IVF outcome data reported annually to the Centers for Disease Control and Prevention (CDC) for the period 2012–2021.

PARTICIPANTS/MATERIALS, SETTING, METHODS

CDC annual reports were used as a data source. Extracted variables included a number of IVF cycle starts, embryo transfers, embryo banking cycles, and live births. IVF ‘treatment efficiency’ was calculated as the proportion of live births per initiated cycle (intent-to-treat analysis).

MAIN RESULTS AND THE ROLE OF CHANCE

Over the study period, IVF cycle starts increased by 234.7% (176 274 to 413 776), embryo transfers by 150.3% (134 471 to 202 121), and embryo banking cycles by 902.3% (18 585 to 167 689). Live births increased by 179.2% (51 286 to 91 906). However, cycle efficiency declined from 29.1% in 2012 to 22.2% in 2021, a relative reduction of 23.4%, with an approximately linear decline across the decade.

LIMITATIONS, REASONS FOR CAUTION

The study is based on aggregate CDC data without patient-level granularity. Therefore, analyses could not control for confounding factors such as parity, infertility diagnosis, or detailed age subgroups. Embryo banking practices also complicate year-to-year comparisons of live birth rates.

WIDER IMPLICATIONS OF THE FINDINGS

These findings suggest that despite substantial growth in IVF cycle numbers, treatment efficiency has progressively declined in the USA over the last decade. This decline coincided with a rapid increase in the use of embryo banking and other IVF ‘add-ons’. The results raise concern about current clinical practices and highlight the need for critical re-evaluation of routine interventions in IVF.

STUDY FUNDING/COMPETING INTEREST(S)

No external funding was received for this study. NG and DHB hold patents related to androgen treatment in females (DHEA) with numbers US8067400B2, US8501718B2, and US9375436B2 (listed in USPTO/public records and assigned to American Infertility of New York). They also receive royalties from Nutraceuticals LLC, which has helped commercialize DHEA as a nutritional supplement.

TRIAL REGISTRATION NUMBER

N/A.

WHAT DOES THIS MEAN FOR PATIENTS?

We analyzed U.S. national data to see how IVF has changed over the past decade and found that, despite a sharp rise in the number of IVF cycles, the chance of having a live birth per cycle in a given year has steadily declined (from about 29% in 2012 to 22% in 2021). Live births increased in absolute numbers, but not as quickly as treatment volume. This downward trend in cycle efficiency coincided with major shifts in IVF practice, including expanded freeze-all strategies and widespread use of related ‘add-ons’. These findings suggest that while IVF is more common, it is becoming less efficient, underscoring the need to re-evaluate routine clinical practices.

Introduction

IVF has been a rapidly growing treatment approach to infertility worldwide. According to the Centers for Disease Control and Prevention (CDC) to which a large majority of U.S. IVF clinics annually report IVF cycle outcomes under a Congressional mandate, IVF cycles increased in the USA in a decade between 2012 and 2021 from 176 274 to 413 776 (a 234.7% increase) (CDC National ART Summary, 2021). For several reasons, this growth has, however, not been uncontroversial. We, for example, already in 2019 reported worldwide declines in IVF live birth rates after approximately 2010 and attributed those at least to a large degree to so-called ‘add-ons’ to IVF (new treatments added after 2010 to IVF protocols) (Gleicher et al., 2019). We then also reported that in the USA—according to CDC data—live birth rates had increased steadily up to 2010 but then started declining till 2016—the at the time last available data set.

There, to the best of our knowledge, have been no follow-up studies on this subject since then, even though changes in IVF practice have been substantial. For example, the utilization of ‘add-ons’ significantly increased further with, often, absolutely no evidence of efficacy. Preimplantation genetic testing for aneuploidy (PGT-A) is a prime example since PGT-A does not improve IVF outcomes, and considerable concern has, indeed, been raised that in certain subgroups of patients, PGT-A may even adversely affect outcomes (Patrizio et al., 2022b; Practice Committees of the American Society for Reproductive Medicine, 2024). Moreover, with the aggressive entry of private equity investments into fertility practice in recent years, because of the field’s growth expectations, IVF has also undergone major structural changes in how and by whom IVF is offered (Patrizio et al., 2022a). And, we recently demonstrated that ownership of IVF clinics affects PGT-A utilization (Patrizio et al., 2025), an observation which, likely, also applies to the utilization of other ‘add-ons’.

Considering all these changes since 2016, we decided to conduct a follow-up study to our 2019 publication, which had ended with the 2016 national U.S. data.

Materials and methods

Ethical approval

This study was based exclusively on publicly available, de-identified aggregate national data reported by the CDC. No individual patient data were used, and therefore, institutional review board approval was not required.

Data source and study design

We again used the annual report of the CDC as a data source, this time presenting data up to the year 2021 (CDC National ART Summary, 2021). Specifically, we used Figure 7 of the document, a table on page 10 out of 19, which offered for the years 2012–2021 (i) number of started IVF cycles, (ii) number of cycles reaching embryo transfer, (iii) number of embryo banking cycles, and (iv) number of live birth deliveries.

Definitions and outcome measure

In CDC terminology, a ‘started IVF cycle’ represents an ovarian stimulation cycle initiated with the intent to retrieve oocytes (intent-to-treat analysis), regardless of whether retrieval or transfer ultimately occurs. This definition includes cycles that result in embryo banking or are cancelled before transfer.

As a measure of annual ‘treatment efficiency’ of IVF, we then calculated the annual live birth percentage for each of the 10 years by determining the percentage of live births in relation to started IVF cycles.

$$\mathit{Cycle} \mathit{efficiency} \left(\%\right)=\left(\mathit{Number} of \mathit{live} \mathit{births} in a \mathit{given} \mathit{year} ÷ \mathit{Number} of \mathit{IVF} \mathit{cycles} \mathit{started} in \mathit{that} \mathit{year}\right)\times 100$$

This efficiency metric, therefore, reflects the probability that a patient starting an IVF cycle in a given year will achieve a live birth within that year. This parameter was chosen to maintain methodological continuity with our earlier national analysis (ending in 2016).

Rationale for efficiency-based analysis

A few points of importance must be made in this context: First, our data are reported with reference cycle start (i.e. intent to treat) because reporting with reference to embryo transfer—as, unfortunately, frequently is done—favorably biases outcomes by basically excluding the least favorable patients from consideration who, in an IVF cycle, fail to produce even one single transferrable embryo. A second important point is to understand why we refer to presented IVF outcomes as a measure of ‘treatment efficiency’ rather than outright live birth rates. Our efficiency measure is distinct from true live birth rates per retrieval or per transfer and is intended as a global indicator of system-level performance under real-world conditions. The reason is best demonstrated in the third column of Table 1, so-called embryo banking cycles: As this column demonstrates, embryo banking cycles increased between 2012 and 2021 from 18 585 to 167 689 (a 902.3% increase). In other words, banking cycles during this decade-long study period increased over four times more rapidly than the above-noted total IVF cycle starts.

Table 1.

Annual ‘cycle efficiency’ of U.S. IVF cycles between 2012 and 2021.

Year	Cycle starts	Embryo transfers	Embryo banking cycles	Live births	Cycle efficiency
	(n)	(n)	(n)	(n)	(%)
2012	176 274	134 471	18 585	51 286	29.1
2013	190 840	137 672	27 564	54 351	28.5
2014	208 630	141 139	35 406	57 340	27.5
2015	231 937	146 590	45 779	60 794	26.2
2016	263 577	154 571	67 106	65 996	25.0
2017	284 385	157 499	87 931	68 908	24.2
2018	306 197	163 889	103 078	73 831	24.1
2019	330 773	171 206	121 086	77 998	23.6
2020	326 468	165 041	123 304	75 023	23.0
2021	413 776	202 121	167 689	91 906	22.2
Increase (%)	+234.7	+150.3	+902.3	+179.2	−23.4

Cycle efficiency (%) was defined as the number of live births reported in a given calendar year divided by the number of IVF cycles initiated in that same year, multiplied by 100. This measure reflects the probability that a patient starting an IVF cycle in a given year will achieve a live birth within that year.

The increase (%) represents the percentage change observed for each outcome over the study period, comparing 2021 to 2014.

This change in IVF practice, therefore, reflects another rapidly growing ‘add-on’ during the study period, which really does not allow for an accurate and unbiased year-over-year comparison of factual birth rates. Because of embryo banking, increasing portions of patients over the study period may not have had an embryo transfer at all in the year of retrieval, and/or may have had a frozen-thawed rather than a fresh embryo transfer during the years.

The assessment of live birth rate per number of started IVF cycles, therefore, does not reflect true live birth rates for a given year but the ‘treatment efficiency’ of an IVF cycle in the USA in a given year to lead to live birth within the same year. It also represents a steady measure applicable to aggregate data as offered by the CDC data set, representing the cumulative utility of all treatments provided in the USA during a started IVF cycle, including all additional ‘add-ons’ in their respective utilization as well as organizational changes in rapidly changing ownership patterns of IVF clinics over the study period. The importance of timely pregnancy achievements for IVF patients has been reported (Roque and Simon, 2020), but is, unfortunately, often overlooked in devising and/or proposing new treatment strategies. We also note, however, that the goal of quick conception must always be balanced with patient safety, including avoidance of ovarian hyperstimulation.

Statistical analyses

Descriptive statistics were used to summarize annual trends in IVF cycle starts, embryo banking cycles, embryo transfers, and live birth deliveries as reported by the CDC. Treatment efficiency was calculated annually as the proportion of live births per started IVF cycle.

Because CDC data are aggregated per initiated cycle and do not separately list the number of ovarian stimulations or cancelled retrievals, a per-stimulation efficiency cannot be directly calculated. We have included this clarification in the Discussion as a methodological limitation.

Results

Table 1 summarizes the reported CDC results.

The results are quite telling: while in the 10-year study period, IVF cycle starts increased by 234.7%, the number of embryo transfers increased by only 150.3%. In other words, U.S. IVF patients lost significant numbers of IVF transfers in this decade, whatever the causes may have been. In parallel, the proportion of oocyte/embryo banking cycles among all ART increased from approximately 10.5% in 2012 to over 40.5% in 2021.

The new ‘add-on’, embryo banking, which increased by a whopping 903.2%, appears to have been a major contributing factor to the decrease in the number of IVF transfers. Considering fewer embryo transfers, it is not surprising that live birth numbers increased less than cycle start (234.7% vs 179.2%), leading to an overall ‘cycle efficiency’ loss of almost a quarter (23.4%) over the decade of the study. This decline followed an approximately linear pattern across the decade, suggesting persistent, cumulative influences from evolving clinical practices.

Discussion

The national IVF cycle outcome data presented here, based on national reporting by U.S. IVF clinics to the CDC over a decade between 2012 and 2021, demonstrate a steady and nearly linear decline in treatment efficiency. This finding extends our previous observation of declining live birth rates after 2010 in the USA and globally (Gleicher et al., 2019).

In practical terms, this means that how IVF has been conducted in the USA over this decade—and likely up to this day—has been increasingly detrimental for patients, even though the median age of patients undergoing IVF has over this study period remained stable at 36 years.

While our analysis cannot establish causality, and what practice changes have been responsible for the further declines after 2016 cannot be determined, several concurrent changes in U.S. IVF practice likely contributed to the observed trend, like the effects of ‘add-ons’, which we raised alarm over for the first time in 2016 (Gleicher et al., 2019). These include the rapid expansion of embryo banking, the increased use of frozen-thawed rather than fresh transfers, wider adoption of PGT-A, extended blastocyst culture, and the growing use of elective single embryo transfer (SET) policies.

These concurrent developments have improved certain safety and quality aspects of IVF (e.g. fewer multiple gestations and higher singleton rates), yet they appear to have collectively contributed to lower treatment efficiency per initiated cycle.

A 903% increase in embryo banking must have significantly contributed to this fall in efficiency, considering the enormity of this increase in utilization, and reports that embryo banking not only fails to improve IVF cycle outcomes, as represented by some of its early proponents (Shapiro et al., 2014), but at least in poor ovarian responders, actually reduces cumulative pregnancy rates for embryo cohorts obtained in a single IVF cycle. Also, as suggested by the data presented here, embryo banking extends the time to live birth (Ge et al., 2023). While embryo banking can reflect valid medical or patient-driven decisions, such as fertility preservation or pre-treatment storage, the magnitude of its increase suggests widespread use in non-medically indicated settings. Unsurprisingly, the exclusion of embryo banking cycles from national IVF cycle reporting, therefore, especially in older women, inappropriately inflated national IVF outcome reporting (Kushnir et al., 2016), and effects of embryo banking, like almost all clinical studies, strongly depend on the general make-up of the study population (Gleicher et al., 2023).

Similarly, extended blastocyst culture and PGT-A, often performed jointly, may disadvantage poor-prognosis patients, particularly those with a low oocyte yield, by reducing the number of transferable embryos available for use, which could potentially have led to viable pregnancies. This reduction further lowers overall treatment efficiency per initiated cycle, as also reported in prior studies (Patrizio et al., 2019).

Even more disturbingly, utilization of useless and often harmful ‘add-ons’ still appears to be increasing in U.S. IVF practice and not only when it comes to such costly ‘add-ons’ as embryo banking (requiring additional treatment cycles) and PGT-A (Patrizio et al., 2022b; Practice Committees of the American Society for Reproductive Medicine, 2024), but also when it comes to issues like separately charging patients for closed incubation time lapse systems. These had already, in 2015, been found not to improve IVF outcomes (Racowsky et al., 2015), as for the longest time claimed by proponents and industry, a finding recently confirmed in a large prospectively randomized study (Kieslinger et al., 2023). More recently, it also involves the ordering of several poorly validated and costly diagnostic tests (Doyle et al., 2022).

Our study has several limitations. It is based on aggregate CDC data without patient-level variables, preventing adjustment for confounders such as parity, infertility diagnosis, or detailed age subgroups. Furthermore, because the CDC aggregates data per initiated cycle, true per-stimulation efficiency cannot be calculated. Consequently, our analysis captures national trends rather than causal mechanisms. Moreover, the inclusion of embryo banking cycles within the total number of ‘started cycles’ likely inflates the denominator, since such cycles often do not involve an immediate embryo transfer, thereby accentuating the apparent decline in annual treatment efficiency.

Despite these limitations, the consistent decline in IVF treatment efficiency across a decade highlights an important signal. While some practice changes have improved safety profiles, others may have inadvertently reduced the overall probability of success for patients starting IVF.

The ultimate result of all of above-cited and many additional changes in clinical IVF practice in the USA since approximately 2010 has, consequently, been a continuous decline in the clinical ‘efficiency’ of IVF practice. What, however, makes these findings especially disappointing is that live birth rates and pregnancy ‘efficiency’ in IVF have been steadily declining, while IVF cycle costs in parallel have been constantly increasing, making IVF increasingly unaffordable and, therefore, inaccessible to the broader public. Although the causes cannot be definitively determined from aggregate data, the trend appears consistent with the cumulative effects of evolving clinical practices, including increased embryo banking, wider use of PGT-A, extended culture, and single-embryo transfer policies.

This study identifies a clear signal rather than proof of causality, but it underscores the need for critical re-evaluation of current IVF strategies to ensure that innovation does not compromise the ultimate goal: timely and successful pregnancy outcomes for patients.

Data availability

The data underlying this article are publicly available from the CDC ART National Summary Reports. All data used in this study are aggregated, de-identified, and can be accessed online at: https://www.cdc.gov/art/php/national-summary/.

Authors’ roles

Concept: NG. Data collection and analysis: NG, SG-L, DHB. Manuscript first draft: NG. Revisions: All authors. Revision after submission and author’s answer to reviewers: LG-B and SG-L. Overall oversight: NG. All authors agreed with the final version.

Funding

No external funding was received for this study.

Conflict of interest

NG and DHB hold patents related to androgen treatment in females (DHEA) with numbers US8067400B2, US8501718B2, and US9375436B2 (listed in USPTO/public records and assigned to American Infertility of New York). They also receive royalties from Nutraceuticals LLC, which has helped commercialize DHEA as a nutritional supplement.