How many is too many?

The finding of highest LBR in high responder group of 16–25 eggs collected can be explained by the fact that there would be more embryos for cryopreservation for subsequent FER than for the low and normal responders. A previous large study1 using the HFEA data from the 1991–2008 period has shown that the LBR increases with number of oocytes up to 15, plateauing between ~15 and 20 oocytes, and steadily decline beyond 20 oocytes.1 However, the authors could not evaluate the LBR rates based on data available at the time. The low LBR in the extremely high responders could be due to poor egg and embryo quality but more probably are associated with high oestradiol concentration secondary to over stimulation.18 19 Furthermore, high ovarian response is often complicated by Polycystic ovary syndrome (PCOS) and can lead to a higher likelihood of freeze-all cycles.20 The proportion of euploid embryos per oocyte yield or stimulation dose remains small.21–23 The larger the number of retrieved oocytes, the higher the number of immature oocytes and the number of oocytes failing to produce an embryo.24 These findings suggest that stimulation protocols appear to rely on data from retrospective studies prompting the ‘more is better’ approach needs changing.

We found a higher LBR in high responders, however, this should not lead to the use of a stimulation regimen that increases oocyte numbers beyond the 15-oocyte mark as it is associated with an increased risk of OHSS. In addition, our data indicate that beyond the 15-oocyte collection mark there is a significant increase in MGP rates, while the LBR declines significantly beyond 25-oocytes (table 3). The biological conversion efficiency from oocytes to achieving an LB potentially drops as the number of retrieved oocytes increases as shown from lower LBR in the very high responders (figure 2). Previous studies have not reported on the MGP rates against oocyte numbers retrieved. In the present study, we show that MGP rates remain similar, but nationally high, in the normal, high and extreme high responders. Despite the widely adopted elective SET (eSET) protocol in women of ≤35 years in the UK, the high MGP rates in the HFEA database suggest that eSET is not widely practised. By contrast, the higher rates of MGP after higher oocytes collected subgroup suggest that the oocyte and subsequent embryo qualities are poor which may explain why many IVF clinics perform multiple ET with MGP rates well above those resulting from spontaneous conception. MGP rates in subgroups where >15 oocyte was collected per cycle oppose the concept that collecting more oocytes will lead to greater embryo choices which improve success rates. The likelihood of double or triple ET for low responder group due to embryo quality and quantity is well known and is associated with the lowest LBR. Our findings suggest that there is a rate limiting factor above which either oocyte competency or embryo qualities are compromised.

Biological and clinical efficiencies

Overall, our data indicate that the level of biological inefficiency of converting oocytes and embryos into an LB is 2.8%. This shows that failure rate is 97.2% from oocytes collected and <5% of all embryos resulting from IVF result in an LB. These data suggest that there is an oocyte competence factor which may compromise embryo quality, and thus LBR. The conversion from oocytes to LB has previously been set at 5%25 whereas in a donor model, the oocyte-to-LBR was found to be 6.5%.26 Compared with pioneering days of IVF, there is evidence that the efficiency of oocyte utilisation rates has declined27 despite the use of numerous non-evidence-based add-on procedures and tests. Unlike previous studies, our contemporary data analysis coincides with a period with add-on practices being used to optimise LBRs and minimise MGP, but these add-ons appear to benefit minimally.

Most patients would likely have undergone ET as there is a 3–6-year post oocyte retrieval time lapse. Previous studies have challenged the concept that the higher the number of oocytes collected per cycle the higher the LBR. Of 7422 women undergoing oocyte retrieval, only 24% conceived with the highest pregnancy rates observed when 13 oocytes were retrieved.16 A previous analysis of the HFEA data from 1991 to 2008 found that an average of 15 oocytes per cycle was optimal for the LBR.1 This large timespan of 1991–2008 was gained from a static database with self-selected cases. Furthermore, IVF practices have changed so much and where MGP and OHSS levels would have been higher than in the 2015–2018 period of our study. Furthermore, the definitions of, for example, infertility and male factor have been in a state of flux, while reporting to the HFEA over the 1991–2008 has at best been variable. By contrast, our study reflects contemporary practices, and our dataset was retrieved from a live database by independent HFEA personnel. In a separate analysis of 14 469 patients from 2009 to 2014, 20 oocytes retrieved appeared optimal for a cumulative LBR, declining thereafter, although >25 oocytes were considered to increase cumulative LBR for ‘freeze-all’ cycles.6 A >16 oocytes retrieved appeared optimal although the study had sparse data beyond 15 oocytes.28 There is some evidence that women with ≥11 oocytes retrieved have compromised cytoplasmic immaturity explaining lowered fertilisation and implantation rates.29 In a mild stimulation protocol on 862 cycles, a 12-oocyte number was found to be optimal for an LB.24 These findings suggest that the widely held concept that having more embryos to choose from needs to be re-evaluated as embryo morphology alone is limited in predicting a successful LB. Although the vast numbers of oocytes and embryos created, it is clear that the biological efficiency to achieving a live birth remains poor at only 4.96% from embryos created, and of 2.84% from oocytes collected (table 2). The dataset does not reveal the level of the freezing oocytes and embryos and its use, but most frozen material is widely expected to be discarded in the long term. For this reason, patients need to be counselled accordingly and to have a realistic expectation of using frozen material especially after 3–6 years of oocyte retrieval and especially for fertility patients not receiving damaging treatments such as with cancers.

Although we found that 16–25 oocytes per retrieval cycle is optimal for an LBR, this was associated with significant increase in the rate of MGP. Neonatal risks in singleton births where excessive ovarian response, that is, when >20 oocytes have profound increased risk of PTB and LBW,8 although unproblematic in another report.17 Our data also inform that there was a significant decline in LBR beyond the 16–25 oocyte retrieval mark suggesting there may be compromising the oocyte at levels unknown yet. We previously found that in 2016 the MGP rate was 10.9% with a direct cost burden for MGP resulting from IVF of £115 million in the UK.14 Milder stimulation protocols and reduced gonadotrophins will likely benefit patients outcomes and also financially. Furthermore, there has been an increase in the numbers of IVF cycles performed as a primary treatment and there is a likelihood that many potentially good responders to IUI are offered IVF as a primary therapeutic approach.14 High oocyte yield is often found in patients with adequate ovarian reserve. Thus, when the male partner presents with adequate semen analysis, milder stimulation protocols or less risky and cost effective IUI cycles should be the primary therapeutic approach.14 24

Future practice and policy outlook

The HFEA did not provide data on OHSS or outcomes for the >50 oocyte subgroup for our study. This was because small numbers could potentially breach confidentiality. It is possible that in the last decade, the switch over to antagonists’ use has reduced the OHSS risks. However, a recent report has shown that severe OHSS and MGP levels remains high despite eSET practices being preferred.14 Cycles with significantly high ovarian response are associated various health risks for both patients and offspring other than OHSS but these data are also lacking. High numbers of oocytes retrieved have impact on embryology laboratory time, staffing needs, cost of additional culture dishes, media used, freezing procedures and maintenance. Furthermore, the added cost associated with FET cycles have never been included in cost effectiveness studies. The cost of freezing excess oocytes and embryos is often absent from cost evaluation as is the added technical costs of FER cycles. The impact of intense egg collections on ovarian biology and future egg quality due to excessive oxidative stress damage to the ovary blood supply and on the health of the patient including pain and bruising, are rarely taken into consideration in most studies. Any benefit of choosing from a greater pool of embryos of having more euploid embryos is negatively balanced by the higher rate of MGP and declining LBR with higher oocyte numbers. In addition, the couple counselling suggesting that the ‘more oocytes is better’ is misleading and psychologically harmful and unethical in generating false hope to achieve a successful pregnancy. Another important consideration relates to extreme >50 oocyte retrieval and the absence of short-term and long-term follow-up information on the health and welfare of these women.

Strengths and limitations

To the best of our knowledge, this study is the largest non-donor aggregate number of IVF treatment cycles showing the association between the number of oocytes retrieved and the LBR and MGP rates, thus reflecting the efficiency of IVF practices. The contemporary period of 2015–2018 also means IVF practices and subfertility definitions are more uniform compared with other studies. The data size provides a reliable baseline information and generalisability on the extent of oocyte retrieval practices and the biological limitations in delivering an LB and enable the evaluation of current clinical management practice. The dataset also provides for baseline failure to retrieve oocytes. The dataset was independently obtained directly from the UK regulatory body thus overcoming selection biases at several levels and ensuring a high degree of applicability of the study results. The 3–6 years post oocyte retrieval follow-up gives a reasonable timeline for fresh or frozen ET to have occurred. While desirable to have a prospective RCT, the size of the dataset is unlikely to match this aggregate dataset analysis.

The main limitations of our study are its retrospective design and the unavailability of associated clinical data such or the segregation of IVF and ICSI data. The patient demographic data were limited, and no data could be obtained on OHSS rates, diagnosis or stimulation methods, use of ICSI, eSET cycles and fresh or frozen IVF cycles, number of oocytes and embryos frozen, numbers and quality of embryos transferred, repeat cycles and use of add-on and selection procedures. The HFEA FOI data are not continuous and therefore limited to calculating adjusted RR using a regression analysis. However, the calculated RR as χ² test showed a significant association between the number of oocytes retrieved and LBR and MGP rates, and this gives a more valid comparison between the groups stratified according to the number of oocytes retrieved.

In summary, the study highlights two risk factors: that an embryo derived from a high oocyte retrieval cohort may not be as beneficial as previously thought, and second, the likelihood of compensating any erosion of embryo quality from high oocyte cohort may tempt practitioners to transfer more embryos thereby heightening the risk of MGP. Clearly, there is a need to review UK ovarian stimulation practices and to adopt milder protocols in most cases given the high level of biological inefficiencies in achieving an LB. Current practices suggest an equivalence of 35 oocytes to achieve one LB and this surely can be improved. Within limits of this study there is a need for practice and policy review to reduce the numbers of oocytes being retrieved in IVF treatments and one which is likely to reduce the maternal and neonatal risks.