Building an evidence base for IVF ‘add-ons’

In countries such as the Netherlands, where the way IVF is practiced is almost entirely determined by health insurance providers, add-ons are not generally available and the debate on their appropriate use is less coloured by commercial considerations. However, in jurisdictions that include a significant private sector, the impact of the report by Spencer et al. (2016)xSpencer et al., 2016. ; See all References)Spencer et al. (2016) and subsequent spin-off media responses were considerable, and in the UK the fertility treatment regulator, the Human Fertility and Embryology Authority (HFEA), decided to act. Together with a range of stakeholders, professional societies and patient advocacy groups, the HFEA developed and published a consensus statement on the use of add-ons (HFEA, 2019). The statement sought to guide both professionals and patients on the principles that should be followed when electing whether or not to utilize an add-on therapy with IVF treatment (https://www.hfea.gov.uk). In this document, fertility care providers are reminded to 'offer fertility treatments ethically' and to adopt a ‘culture change’ that, it is implied, protects patients from potential exploitation.

Supporting its publication in the UK, the HFEA launched a patient-orientated website (https://www.hfea.gov.uk/treatments/explore-all-treatments/treatment-add-ons/) that uses a traffic light system to summarise their view of the evidence supporting the safety and efficiency of a selected number of adjuvant laboratory and clinical procedures usually offered at additional cost. Any ‘add-on’ given a green light requires clear supportive evidence from at least one randomized controlled trial. An amber light indicates that the evidence is less clear, whereas a red light warns that no evidence of safety or effectiveness exists to support the use of the treatment in question (Table 1Table 1).

Table 1Human Fertility and Embryology Authority traffic light rating system for IVF ‘add-ons’

Sign	Explanation	Add-ons
Red	No evidence that this add-on is effective and safe	• Assisted hatching; • PGS; • IMSI; • PICSI; • Intrauterine culture; • Reproductive immunology tests and treatment
Amber	There is a conflicting body of evidence for this add-on, further research is required	• Artificial egg activation calcium ionophore; • Elective freeze all cycles; • Embryo glue; • Endometrial scratching; • PGS; • Time-lapse imaging
Green	RCTs prove that the add-on is effective and safe	None

View Table in HTML

RCT = randomized controlled trial.

Adapted from www.hfea.gov.uk/treatments/explore-all-treatments/treatment-add-ons.

Within the context of private sector-dominated practice, such as that in the UK, there is much to welcome in this initiative as it appears to meet the need to inform patients without completely denying patient-access to potentially beneficial innovations. Indeed, this model might also be applicable in other countries with strong commercial IVF sectors. It is however salutary to note that after 40 years of IVF, none of the interventions commonly used in practice are underpinned by the evidence base required by the HFEA to qualify for a green light. The orthodox response to this research gap has long been, and remains, to fill it with ‘more randomized controlled trials’ (Repping, 2019xRepping, 2019. ; : See all ReferencesRepping, 2019). In the meantime, patients and clinicians are still left stranded: chided for using unproven interventions, but not provided with evidence supporting alternative approaches.

In the current evidence-based medicine paradigm, randomized controlled trials (RCTs) remain at the top of the hierarchy of evidence quality, only surmounted by meta-analysis (the combined analysis of RCTs) and, even better, individual patient data (IPD) meta-analysis. The case for this can be strongly made, as it would appear to be the only means of removing potential biases and confounding factors that can otherwise invalidate research findings. So, policy makers consider that in order to discern what can be ‘ethically’ offered to patients, the only valid evidence must come from RCTs (hfea.gov.uk). We may all be in for a long wait.

Why? Firstly, RCTs have become very hard to do (Humaidan and Haahr, 2019xHumaidan and Haahr, 2019. ; ()See all ReferencesHumaidan and Haahr, 2019). Study proposals are subject to intense and prolonged regulatory scrutiny from multiple stakeholders, costing both time and money. If ultimately approved, implementation of the study often requires the marshalling of multiple participating centres and a monitoring infrastructure that is so burdensome that only few research groups can access the resources required.

In order to carry out a definitive RCT with sufficient power to demonstrate a clinically and statistically significant impact on live birth (the only outcome now acceptable to many reviewers), more than 1500 randomized subjects are often required (Stocking et al., 2019xStocking et al., 2019 ; ()See all ReferencesStocking et al., 2019). Most grant applications for such studies must therefore seek funding in excess of €1,000,000. And once the necessary finance and recruitment infrastructure is secured, the challenges of convincing the often frustratingly few patients who meet the tight inclusion criteria to participate become apparent. Few well-planned and well-funded studies will meet their recruitment target, and those that do achieve this because of the remarkable commitment and resilience of the investigators, and often at the cost of additional funding.

Given the time required between inception and publication of a high quality RCT, the long- awaited results may, on publication, be of little relevance to contemporary practice. Examples of this include the ‘ANTARCTICA’ trial of endometrial preparation methods for frozen-thawed embryo transfer (Groenewoud et al., 2016xGroenewoud et al., 2016 ; : See all ReferencesGroenewoud et al., 2016) and the ‘ESTEEM’ trial of preimplantation genetic screening using the polar body (Verpoest et al., 2018xVerpoest et al., 2018 ; : ()See all ReferencesVerpoest et al., 2018). Both were designed and powered to represent a high quality RCT but were criticised at publication because the technologies employed were now considered obsolete. Both studies had taken several years longer than anticipated to perform and yet, despite careful design and considerable effort, their impact on practice risks being negligible.

It is not just the technologies that evolve. The perceived needs and interests of our patients change too. Examples of treatment approaches that were introduced to meet perceived patient demand that changed by the time an acceptable evidence base had been generated include egg sharing (Ahuja et al., 1999xAhuja et al., 1999 ; : See all ReferencesAhuja et al., 1999), now increasingly replaced by egg banking, and mild stimulation approaches (Nargund et al., 2017xNargund et al., 2017 ; : See all ReferencesNargund et al., 2017), the espoused safety advantages of which are now widely considered to be met by GnRH agonist triggering, freeze-all cycles and single embryo transfer. Other criticisms of RCTs that have been well rehearsed elsewhere include the risk that imposing narrow inclusion criteria will limit the applicability of the findings to a broader patient population, whereas permitting broader inclusion criteria means individual variation in response to the studied intervention may be so great as to mask any effect.

Despite these challenges, major and laudable RCTs addressing clinical questions in our field reach publication in top journals. However, in addition to sharing the necessary major financial and manpower investment to perform, their clear tendency to produce negative findings means that they are primarily serving to remove treatment options from the clinician and their patient. This can of course represent an important contribution. However, when such trials test empirical treatments (which many IVF ‘add-ons’ are), they risk increasing confusion rather than clarity. Recent examples include the ‘PROMISE’ trial which concluded that progesterone supplementation is ineffective in the treatment of recurrent miscarriage (Coomaraswamy et al., 2015xCoomarasamy et al., 2015 ; : See all ReferencesCoomaraswamy et al., 2015), and a major trial that declared endometrial scratching useless for improving IVF outcomes (Lensen et al., 2019xLensen et al., 2019 ; : See all ReferencesLensen et al., 2019). Both of these studies represent considerable achievements, but their conclusions must be challenged because instead of investigating the effectiveness of an intervention designed to address a specific pathology, they have studied the empirical treatment of a clinical presentation. Recurrent miscarriage and implantation failure are terms used to describe experiences that have a range of aetiologies. It is therefore possible that blindly applied empirical therapy may have negative effects in some participants and positive effects in others, resulting in no overall impact being discerned. So, while some participants may have benefited, the intervention is declared worthless and complete rejection is prematurely recommended.

Recognition of some of the weaknesses of the RCT model has led to the emergence of individual patient data analysis (IPD). While this represents an important means of adding value to the investment in RCTs, IPD addresses only some of the challenges outlined above. RCTs remain a science of detecting mean rather than individual differences, and the nascent era of personalized medicine calls for a different approach.

Rather than requiring RCTs to discern the value of IVF ‘add-ons’, we propose that the research and clinical communities come together in a new and different collective enterprise: one that can harness rather than suppress the drive to introduce innovations that our patients often lead. And one that generates the data to guide clinical practice that RCTs have largely failed to do.

The huge growth in bioinformatics and data management that is transforming every sector is now ripe for application in reproductive medicine. Clinics can and should work together to develop large prospective carefully phenotyped cohorts, using an expanding array of validated diagnostics as well as clinical and demographic data to deeply characterize individual phenotypes. Prospective collection of carefully defined outcome criteria, close follow-up and reliable archiving of data are disciplines that can be taught by those expert in running conventional multicentre studies. If this model can attract similar funding as ‘blockbuster’ RCTs, it offers a feasible and scientifically sound alternative means of determining the real world value of novel innovations.

Such an initiative would radically differ from current evidence-based research orthodoxy. All patients could be included, rather than just a sample. The focus would be on the individual patient characteristics that affect outcomes of a given intervention, rather than the overall effect of the intervention(s) in a study population. The opportunities for generating large cohorts offered by increasing consolidation in the IVF market are enhanced by the maturing of techniques of big data analysis. Guidelines such as the 'TRIPOD' statement can assist consistent approaches to multivariate analysis (Moons et al., 2015xMoons et al., 2015 ; : See all ReferencesMoons et al., 2015) and powerful algorithms can be developed that can generate the data and advice pertinent to individual patients. Rather than avoiding the impact of individual confounding variables, this approach would embrace them, meeting the needs of the current era of personalised medicine (Christ et al., 2019xChrist et al., 2019 ; : See all References, Mega et al., 2014xMega et al., 2014. ; : See all References).

Such a change in paradigm would open a way for national ART regulators to contribute to the generation of clinical research data and not just to its assessment. In many countries, such as the UK, prospective collection of individual cycle data and outcomes is mandatory. The core infrastructure for such an initiative therefore exists. Reviewing and extending the parameters that are recorded to include the diagnostic and clinical elements employed in practice would enable the link to be made between the underlying diagnosis and the effectiveness of therapies for specific patient phenotypes. As the HFEA seeks to change the culture in IVF practice in relation to the use of add-on therapies, it could require clinics that offer these interventions to register their use alongside established data elements including clinical outcomes.

Large, rigorous RCTs that reach publication will continue to be of great value to our field (Wei et al., 2019xWei et al., 2019. ; ()See all ReferencesWei et al., 2019), but to consider them the sole arbiters of ‘clinical truth’ risks leaving many urgent clinical questions unanswered. Accepting their limitations (Amrhein et al., 2019xAmrhein et al., 2019. ; : See all References, Stocking et al., 2019xStocking et al., 2019 ; ()See all References) and the potential value of alternative approaches would represent an important step towards providing the information needed to guide clinicians in the appropriate use of clinical interventions that conventional evidence-based research orthodoxy has largely failed to deliver.