A Cost-Effectiveness Analysis of Preimplantation Genetic … · 3.4.4.2 Incremental Costs: Health States.....25 3.4.4.3 Markov Model Transition Costs.....25 ... RPL – recurrent

A Cost-Effectiveness Analysis of Preimplantation Genetic Screening with In Vitro Fertilization In Ontario

by

Rhonda Gina Zwingerman

A thesis submitted in conformity with the requirements for the degree of Master of Science

Institute of Health Policy, Management & Evaluation,

University of Toronto

© Copyright by Rhonda Gina Zwingerman 2017

ii

A Cost-Effectiveness Analysis of Preimplantation Genetic Screening

with In Vitro Fertilization in Ontario

Rhonda Gina Zwingerman

Master of Science

Institute of Health Policy, Management & Evaluation

University of Toronto

2017

Abstract

Background: Preimplantation genetic screening (PGS) is an increasingly common addition to IVF

cycles in an effort to improve upon current embryo selection techniques by limiting the transferrable

cohort of embryos to those that are euploid. Objective: To examine the cost-effectiveness of

preimplantation genetic screening (PGS) with IVF compared to IVF alone in terms of additional live

births. Methods: A decision analytic model was created using TreeAge Pro to compare IVF with

PGS to IVF alone for women of different age groups with one to ten blastocysts from the societal

perspective. Sequential single embryo transfers of blastocysts from a single oocyte retrieval were

modelled until all blastocysts were exhausted, treatment was discontinued or a live birth was

achieved. Results: PGS became both incrementally less costly and more effective as the number of

blastocysts and the age of the woman increased. IVF alone was the superior strategy – more effective

and less costly - for all women under age 35. This was also true when three or fewer blastocysts were

available, regardless of age. Sensitivity analysis showed that the model was most sensitive to the

implantation rates and early pregnancy loss rates of both IVF with PGS and IVF alone. Conclusions:

Using currently available data, PGS appears to be more costly and less effective than IVF alone in

the majority of scenarios. However, PGS may be less costly and more effective for older women with

a large number of blastocysts. Future research should focus on improving available population-based

PGS outcome data.

iii

Dedicated to Benjamin.

iv

Acknowledgments

I would like to wholeheartedly thank, first and foremost, my thesis committee, Dr. Okun, Dr. Shapiro

and Dr. Isaranuwatchai, for their mentorship, guidance, commitment and support throughout this

entire process. Without each of them – both individually and collectively – this work would not have

been possible. I would also like to sincerely thank my internal and external reviewers, Dr. Kolomietz

and Dr. Hitkari, for taking the time out of their very busy schedules to help me realize this goal. I

would like to express my gratitude to all the faculty and students at IHPME who taught, supported

and guided me along this journey and widened my field of vision beyond what is taught in medical

school and residency.

None of this would have been possible without the continued support and encouragement of my

friends, family and, of course, my husband, Jeremy. Thank you all from the bottom of my heart. I

could not have done this without you.

I would like to thank and acknowledge BORN Ontario for access to CARTR Plus / BORN Ontario

data and the Department of Obstetrics and Gynaecology, Mount Sinai Hospital, Toronto for their

generous funding support.

v

Table of Contents

Acknowledgments.......................................................................................................................... iv

Table of Contents .............................................................................................................................v

List of Tables ................................................................................................................................. ix

List of Figures ..................................................................................................................................x

List of Abbreviations ..................................................................................................................... xi

1 Introduction .................................................................................................................................1

1.1 Infertility in Canada .............................................................................................................1

1.2 In Vitro Fertilization ............................................................................................................1

1.3 Female Age & Embryo Aneuploidy ....................................................................................2

1.4 The Rationale behind Preimplantation Genetic Screening ..................................................2

1.4.1 Improving IVF Outcomes ........................................................................................2

1.4.2 Elective Single Embryo Transfer .............................................................................3

1.5 The Rationale behind Economic Evaluations ......................................................................3

2 Review of the Literature..............................................................................................................7

2.1 A Brief History of PGS ........................................................................................................7

2.1.1 Original Methods of Embryo Biopsy & Aneuploidy Screening ..............................7

2.1.2 Trophectoderm Biopsy.............................................................................................8

2.1.3 Comprehensive Chromosome Screening .................................................................9

2.1.3.1 Array Comparative Genomic Hybridization .............................................9

2.1.3.2 Single Nucleotide Polymorphism Analysis ...............................................9

2.1.3.3 Real-Time Quantitative PCR ...................................................................10

2.1.3.4 Next Generation Sequencing ...................................................................10

2.2 The Efficacy of Comprehensive Chromosome Screening .................................................10

2.3 The Cost-Effectiveness of PGS .........................................................................................11

2.3.1 CEAs of Older Methods of PGS ............................................................................11

vi

2.3.2 CEAs of PGS for Recurrent Pregnancy Loss ........................................................12

2.3.3 CEAs of PGS using CCS .......................................................................................12

2.4 Study Objective ..................................................................................................................14

3 Methods .....................................................................................................................................16

3.1 Study Design ......................................................................................................................16

3.2 Model Design & Structure .................................................................................................16

3.3 Clinical Parameters ............................................................................................................17

3.3.1 CARTR Plus / BORN Ontario ...............................................................................18

3.3.2 Pregnancy & Pregnancy Loss Rates ......................................................................19

3.3.3 Treatment Discontinuation.....................................................................................19

3.4 Cost Parameters – Societal Perspective .............................................................................20

3.4.1 Assisted Reproductive Technologies & Medication Costs ....................................20

3.4.2 Productivity Costs ..................................................................................................21

3.4.3 Hospital and Physician Costs .................................................................................22

3.4.4 Calculating Costs within the Decision Tree...........................................................23

3.4.4.1 Initial Costs: Health States ......................................................................23

3.4.4.2 Incremental Costs: Health States .............................................................25

3.4.4.3 Markov Model Transition Costs ..............................................................25

3.5 Model Assumptions ...........................................................................................................27

3.6 Alternative Perspectives.....................................................................................................28

3.7 Sensitivity Analyses ...........................................................................................................29

3.7.1 One-way Sensitivity Analysis ................................................................................29

3.7.2 Probabilistic Sensitivity Analysis ..........................................................................30

3.8 Ethics Approval .................................................................................................................30

4 Results .......................................................................................................................................31

4.1 Societal Perspective ...........................................................................................................31

vii

4.1.1 Incremental Cost ....................................................................................................31

4.1.2 Incremental Effect ..................................................................................................31

4.1.3 Incremental Cost-Effectiveness Ratios ..................................................................32

4.1.4 Sensitivity Analyses ...............................................................................................32

4.1.4.1 One Way Sensitivity Analysis .................................................................32

4.1.4.2 Probabilistic Sensitivity Analysis ............................................................33

4.2 Alternative Perspectives.....................................................................................................34

4.2.1 Direct Health Care .................................................................................................34

4.2.2 Hypothetical MOHLTC .........................................................................................35

4.2.3 Patient Self-Pay ......................................................................................................35

5 Discussion .................................................................................................................................36

5.1 Key Findings ......................................................................................................................36

5.2 Comparison with Other Studies .........................................................................................38

5.3 Strengths ............................................................................................................................40

5.4 Limitations .........................................................................................................................41

5.4.1 Clinical Parameters ................................................................................................41

5.4.2 Cost Data ................................................................................................................42

5.4.3 Model Design .........................................................................................................43

5.5 CARTR Plus / BORN Ontario ...........................................................................................43

5.5.1 Data Quality ...........................................................................................................44

5.5.2 Data Accessibility ..................................................................................................47

5.6 Non-Economic Advantages of PGS...................................................................................48

5.7 Future Directions ...............................................................................................................49

5.8 Conclusions ........................................................................................................................50

6 References .................................................................................................................................51

7 Tables ........................................................................................................................................62

viii

7.1 Description of clinical inputs included in the model. ........................................................62

7.2 Detailed description of all costs included in the model. ....................................................63

7.3 Costs included under the four different perspectives examined. .......................................66

7.4 Point estimates for C, E and ICER by age category and number of blastocysts

(societal perspective). ........................................................................................................68

7.5 Results of the probabilistic sensitivity analysis by age category and number of

blastocysts (societal perspective). ......................................................................................69

7.6 Point estimates for C, E and ICER by age category and number of blastocysts

(alternative perspectives). ..................................................................................................70

7.7 Causes of inadequate PGS data quality and suggestions for data quality improvement

in the CARTR Plus / BORN Ontario registry. ...................................................................73

8 Figures .......................................................................................................................................74

8.1 The cost-effectiveness plane. .............................................................................................74

8.2 Simplified diagram of the Markov model used within the IVF alone arm of the full

decision tree. ......................................................................................................................75

8.3 Simplified diagram of the IVF/PGS arm of the decision model, including the

embedded Markov model and its clones. ...........................................................................76

8.4 Full decision tree comparing IVF/PGS to IVF alone. ........................................................77

8.5 One-way sensitivity analysis: tornado diagrams by age group and number of

blastocysts, from a societal perspective. ............................................................................78

8.6 Probabilistic sensitivity analysis: scatter plots by age group, from a societal

perspective. ........................................................................................................................80

8.7 Probabilistic sensitivity analysis: scatter plots by age group, from alternative

perspectives. .......................................................................................................................81

8.8 Components of data usability.............................................................................................83

9 Appendices ................................................................................................................................84

9.1 Categorization of causes of inadequate data quality in medical registries. .......................84

ix

List of Tables

1. Description of clinical inputs included in the model.

2. Detailed description of all costs included in the model.

3. Costs included under the four different perspectives examined.

4. Point estimates for C, E and ICER by age category and number of blastocysts (societal

perspective).

5. Results of the probabilistic sensitivity analysis by age category and number of blastocysts

(societal perspective).

6. Point estimates for C, E and ICER by age category and number of blastocysts

(alternative perspectives).

7. Causes of inadequate PGS data quality and suggestions for data quality improvement in

the CARTR Plus / BORN Ontario registry.

x

List of Figures

1. The cost-effectiveness plane.

2. Simplified diagram of the Markov model used within the IVF alone arm of the full

decision tree.

3. Simplified diagram of the IVF/PGS arm of the decision model, including the embedded

Markov model and its clones.

4. Full decision tree comparing IVF/PGS to IVF alone.

5. One-way sensitivity analysis: tornado diagrams by age group and number of blastocysts,

from a societal perspective.

6. Probabilistic sensitivity analysis: scatter plots by age group, from a societal perspective.

7. Probabilistic sensitivity analysis: scatter plots by age group, from alternative

perspectives.

8. Components of data usability.

xi

List of Abbreviations

aCGH – array comparative genomic hybridization

ARA – advanced reproductive age

BORN – Better Outcomes Registry & Network

CARTR – Canadian Assisted Reproductive Technologies Register

CCS – comprehensive chromosome screening

CEA – cost-effectiveness analysis

CGH – comparative genomic hybridization

CLBR – cumulative live birth rate

COH – controlled ovarian hyperstimulation

DET – double embryo transfer

EPL – early pregnancy loss (used here in lieu of the terms miscarriage or spontaneous abortion)

eSET – elective single embryo transfer

ET – embryo transfer

FET – frozen embryo transfer

FISH – fluorescent in situ hybridization

ICER – incremental cost-effectiveness ratio

ICSI – intracytoplasmic sperm injection

IVF – in vitro fertilization

LBR – live birth rate

MOHLTC – Ministry of Health and Long Term Care [of Ontario]

NGS – next generation sequencing

OCCI – Ontario Case Costing Initiative

OPU – ovum pickup (a.k.a. oocyte retrieval)

PGD – preimplantation genetic diagnosis

PGS – preimplantation genetic screening

PSA – probabilistic sensitivity analysis

qPCR – quantitative real-time polymerase chain reaction

RPL – recurrent pregnancy loss

SART – Society of Assisted Reproductive Technology

SNP – single nucleotide polymorphism

WTP – willingness-to-pay

1

1 Introduction

1.1 Infertility in Canada

The prevalence of infertility in Canada has been increasing over time, with recent studies estimating

that up to sixteen percent of Canadian couples are affected (1). The inability to conceive is associated

not only with significant emotional and psychological hardship but is all too often accompanied by

social stigma and relationship conflict as well (2). And while treatment options for infertility have

increased both in number and sophistication over the past decades, those treatments are often

stressful, invasive, disruptive and expensive (2).

1.2 In Vitro Fertilization

In vitro fertilization (IVF) was pioneered in the 1970s as a treatment for tubal disease, with the first

live birth occurring in England in 1978 (3). In its original incarnation, IVF involved the laparoscopic

retrieval of an oocyte from a single dominant follicle, the fertilization of that oocyte by sperm in

vitro, the culture of the resulting embryo to the cleavage stage of development and the transfer of that

embryo back into a woman’s uterus in the hopes of achieving a pregnancy (3). Since 1978, the

process as well as the indications for IVF have evolved significantly. IVF is now used to help people

struggling with infertility due to almost any cause (4). It is also used to help single and LGBTQ

individuals and couples build their families. Modern IVF begins with controlled ovarian

hyperstimulation (COH) with injectable gonadotropins to encourage the development of multiple

mature oocytes. The oocytes are then retrieved transvaginally under ultrasound guidance – a

procedure that is often referred to as an oocyte pick-up (OPU). Those oocytes are then fertilized

either traditionally, where tens of thousands of sperm are incubated with each oocyte and fertilization

is allowed to occur on its own, or by intracytoplasmic sperm injection (ICSI), where an individual

sperm is injected into each oocyte. The resultant embryos are then cultured to the blastocyst stage,

which, in humans, is reached five to six days after fertilization. Blastocysts are either transferred into

a woman’s uterus or cryopreserved for future use. Since 1978, millions of babies worldwide have

been born through IVF (4). Despite these advances, the overall success rates of IVF remain low. In

Canada in 2014, the chance of live birth per cycle started was only 22.3% (5).

2

1.3 Female Age & Embryo Aneuploidy

Aneuploid embryos are embryos that have an incorrect number of chromosomes. Conversely,

euploid embryos are those that have the correct chromosomal complement. In humans, euploid cells

have 22 pairs of autosomes plus 2 sex chromosomes. Human gametes, oocytes and sperm, each have

half the euploid number of chromosomes – 22 autosomes and one sex chromosome.

It is now well recognized that incidence of aneuploidy in gametes and embryos is high and that

aneuploidy is a significant cause of the relatively low fecundability in humans (6,7). Aneuploidy

rates in human oocyte and embryos are strongly correlated with oocyte age, and therefore, female

age. Increasing aneuploidy rates with female age parallel a rise in spontaneous early pregnancy loss

(EPL) rates and a decline in natural fertility (4). Studies have shown fewer than 30% of blastocysts

created for use in IVF are aneuploid in women aged 26 to 30 but that this number rises to

approximately 85% for women over 43 years of age (8).

1.4 The Rationale behind Preimplantation Genetic Screening

1.4.1 Improving IVF Outcomes

Currently, morphological characteristics of embryos are used to select the best embryo(s) for transfer

in an IVF cycle. While there does exist a correlation between embryo morphology and chromosomal

status, with higher quality blastocysts more likely to be euploid, this correlation is moderate at best

(9). Many high quality blastocysts are in fact aneuploid, particularly in women of advanced

reproductive age (ARA) (9). The transfer of aneuploid embryos is a significant cause of failure to

achieve pregnancy (implantation failure) and of EPL after IVF (10). For women or couples with

multiple embryos available to transfer, the repeated transfer of embryos that unknowingly lack the

capacity to lead to a live birth is not only time consuming and costly, but also psychologically

difficult. This is particularly true when those embryos result in nonviable pregnancies.

Preimplantation genetic screening (PGS) is a process whereby embryos that are created in vitro are

biopsied in order to determine their chromosomal status. That information is then used to select the

most appropriate embryo or embryos for transfer rather than relying on embryo morphology alone.

Because transferred aneuploid embryos lead to implantation failure or EPL, it follows that restricting

transferred embryos to those that are euploid should improve implantation and live birth rates (LBR)

per embryo transferred (10,11).

3

1.4.2 Elective Single Embryo Transfer

In addition to improving embryo selection, PGS also has the potential to increase the acceptability of

elective single embryo transfer (eSET). Because of low implantation rates per embryo transferred, it

has long been accepted practice to transfer multiple embryos at a time. This practice has resulted in

very high rates of twins and higher order multiples for those undergoing IVF and consequently

unacceptably high rates of obstetrical and neonatal complications associated with multiple

pregnancies, most notably the complications associated with prematurity. More recently there has

been a movement to promote increased use of eSET, particularly in younger women with favourable

chances of success (12,13). And while some jurisdictions have tied IVF funding to eSET policies,

these have applied almost exclusively to younger women, with double embryo transfer (DET)

policies remaining the standard for many (14,15). With the improved implantation rates of confirmed

euploid embryos, physicians and patients alike may be more willing to practice eSET in

circumstances where DET is the current standard, including in women of ARA (16). In fact, the

American Society of Reproductive Medicine recently released a statement recommending that all

euploid blastocyst transfers be single embryo transfers regardless of the age of the woman (17).

1.5 The Rationale behind Economic Evaluations

In order for new treatments or technologies to be funded by insurance companies or governmental

health ministries a number of factors must be considered. These include the clinical effectiveness of

the proposed new intervention, the budgetary impact of its implementation, potential political

ramifications, and its cost-effectiveness. Although the cost-effectiveness of a given intervention is

often portrayed as an intrinsic quality of that intervention, it is more accurately conceptualized as a

relative construct (18), meaning it is better judged in relation to other available interventions rather

than in isolation. A new intervention may be considered cost-effective by decision-makers in

comparison to current treatment X but not in comparison to alternative treatment Y. A novel health

technology is often compared to the current standard of care to determine its cost-effectiveness under

a specific set of circumstances. Economic evaluations, therefore, are methodological attempt to

quantify the trade-off between those costs and health effects (19–21). A cost-effectiveness analysis

(CEA) is a type of economic evaluation where the health effects are quantified in terms of natural

health units gained rather than in quality-adjusted life years (as in a cost-utility analysis) or dollars

(as in a cost-benefit analysis) (19,22,23).

4

There are two common summary statistics used to report CEA results in medicine: the average cost-

effectiveness ratio (ACER) and the incremental cost-effectiveness ratio (ICER) (23,24). The ACER

is calculated as the average cost of a treatment divided by the average effect; therefore, a comparison

of current practice to a new intervention yields two ACERs which are then compared to each other to

determine their relative cost-effectiveness.

𝐴𝐶𝐸𝑅 = 𝐶𝑎𝑣𝑔

𝐸𝑎𝑣𝑔

where Cavg = the average cost of an intervention; Eavg = the average effectiveness of an intervention

The ICER, on the other hand, is calculated as the difference in costs between the two interventions

divided by the difference in effects between the two interventions.

𝐼𝐶𝐸𝑅 =𝐶𝑛𝑒𝑤 − 𝐶𝑠𝑡𝑑

𝐸𝑛𝑒𝑤 − 𝐸𝑠𝑡𝑑=

𝐶

𝐸

where Cnew = the cost of the new intervention; Enew = the effectiveness of the new intervention; Cstd = the cost of

the standard intervention; Estd = the effectiveness of the standard intervention

While the ACER may be simpler to understand and is commonly reported in publications, it can lead

to inappropriate conclusions. The ACER obscures what is happening at the margins because it

“distributes the difference in costs over all subjects and assumes all outcomes are produced at

equivalent costs” (23). It also does not explicitly compare one intervention to another. For these

reasons, it has been recommended that the ICER be the primary summary statistic used in CEAs

(23).

ICERs are easiest to understand diagrammatically in the context of the cost-effectiveness plane

(Figure 8.1). This is a plane divided into four quadrants through the origin with E on the x-axis and

C on the y-axis (19,22). ICERs that fall within each of the four planes can be interpreted as follows:

1. Northwest quadrant: Here the new intervention is dominated by the current standard of care.

This is because the new intervention is both less effective and more expensive than the

current standard. In this document, this quadrant will be shaded red.

2. Northeast quadrant: Here the new intervention is more effective but also more expensive than

the current standard of care. Many new health technologies and interventions fall into this

quadrant. The ICER is most useful in this quadrant. Ultimately, whether or not a new

intervention with an ICER in the northeast quadrant is deemed to be cost-effective depends

5

on decision-makers’ willingness-to-pay for additional health gains. In this document, this

quadrant will be shaded blue.

3. Southwest quadrant: Here the new intervention is both less effective and less expensive than

the current standard of care. ICERs in this quadrant appear positive because both the E and

C are negative. Unlike the northeast quadrant, it is more difficult to interpret ICERs in this

quadrant. It is also more difficult for decision-makers to consider policies that would result in

worse health outcomes for the sake of financial savings. In this document, this quadrant will

be shaded yellow.

4. Southeast quadrant: Here the new intervention is said to be dominant over the current

standard of care. This is because the new intervention is both more effective and less

expensive than the current standard. Interventions in this quadrant should be readily adopted

into clinical practice. In this document, this quadrant will be shaded green.

An ICER is most helpful for conveying the results of a CEA when the new intervention is more

effective and more costly than the current standard of care, in other words, when the ICER falls in

the northeast quadrant of the cost-effectiveness plane (Figure 8.1). Negative ICERs, i.e. those in the

northwest and southeast quadrants, are not only conceptually difficult to understand but difficult to

analyze. Because the ICER is a ratio statistic, it is not possible to know whether a negative ICER in

isolation is negative because of a negative numerator (C) or because of a negative denominator

(E) (25). Furthermore, ICERs made of up both negative numerators and denominators, i.e. those in

the southwest quadrant, will appear positive. This is problematic as their interpretation is entirely

different from a positive ICER made up of both a positive numerator and denominator. ICERs in the

southwest quadrant indicate that the new treatment is both less expensive and less effective than the

current treatment. The societal pressures and values involved in deciding to adopt a new strategy that

will save money at the expense of health are vastly different than those involved in deciding to spend

more money in order to drive better outcomes.

A series of ICERs all residing in the northeast quadrant of the cost-effectiveness plane can be ordered

in a meaningful way from least to most cost-effective to allow for a relatively intuitive comparison of

options by decision makers. Unfortunately, this is not the case for negative ICERs which cannot be

meaningfully ordered (25). ICERs also become less helpful statistics as the E approaches zero.

Because the ICER is a ratio with E as the denominator, as E approaches zero, the ICER

6

approaches infinity and its absolute value because less useful. When the difference in effectiveness

between the two treatment options is very small or absent, one could choose to simply compare the

costs, although it is preferable to still consider both costs and effects and then employ sensitivity

analysis techniques to characterize the uncertainty around the estimates (19,26).

The ICER, however, is only one piece of information required to determine cost-effectiveness (27).

Ultimately both the budget and the willingness-to-pay (WTP) for additional units of health gained

must be considered as well. The choice of a WTP threshold, or the ICER value above which a

decision maker deems it worthwhile to adopt a given intervention, depends not only on who the

decision maker is but also on how they value particular health gains as well as their risk tolerance for

adverse health outcomes (27). In the case of the current CEA, the unit of health under consideration

is one live birth. Establishing a WTP for this health outcome is complex and remains a pursuit in its

methodological infancy. The theoretical and practical issues surrounding the valuation of an

additional life created as the result of a given treatment is beyond the scope of this paper.

Alternatively, one may conceptualize the outcome of live birth as a successful “cure” of a case of

infertility or childlessness. Despite the fact that the burden of infertility as a disease is known to be

significant and similar to that of other serious chronic diseases (28), there is no agreed upon WTP

threshold, either in the published IVF literature or elsewhere, at which an intervention should clearly

be adopted in order to alleviate this burden. And even if there did exist a consensus WTP threshold,

whether or not an intervention is deemed to be good value for money is only one aspect to consider

when deciding how to allocate limited health care resources. Attention must also be paid to factors

such as equity of access to care, ethics and legislative constraints (27). By their nature, these non-

monetary considerations are arguably more relevant to assisted reproductive technologies compared

to other interventions such as pharmaceuticals or novel surgical approaches.

7

2 Review of the Literature

2.1 A Brief History of PGS

The earliest experiences with the testing of in vitro human embryos for clinical use was for sex

determination for the purposes of avoiding the transmission of X-linked disorders (29,30). This arose

as an alternative to prenatal testing and pregnancy termination for couples at risk of having an

affected child. From these first reports the field of preimplantation genetic diagnosis (PGD) was

born. And as polymerase chain reaction (PCR) techniques improved, the indications for PGD grew

beyond sex determination to include autosomal single-gene disorders and much more, including

translocations and HLA typing (31).

In the early 1990s, as IVF was establishing itself as a mainstream treatment for infertility, experts

began to wonder if PGD could have applications for couples without known genetic diseases but with

anticipated poor outcomes after IVF (6). Knowing that many failed IVF cycles were due to

aneuploidy, PGD techniques were adapted to screen for monosomies and trisomies (32). PGD for

aneuploidy screening was thus born – something that we now refer to as PGS. PGS screens embryos

for aneuploidies when both genetic parents are known or presumed to be chromosomally

normal whereas PGD tests embryos at risk for the presence or absence of a specific genetic

mutation or chromosomal rearrangement (33).

2.1.1 Original Methods of Embryo Biopsy & Aneuploidy Screening

Genetic material for PGD or PGS can be obtained with one of three techniques: a polar body biopsy,

a blastomere biopsy at the cleavage stage or a trophectoderm biopsy at the blastocyst stage (33,34).

Until relatively recently, fluorescence in situ hybridization (FISH) was used to screen biopsy samples

for aneuploidies. FISH is a technique that uses multiple coloured fluorescent probes that bind to

DNA sequences unique to each chromosome (33). The number of different probes used determines

the number of chromosomes screened for and therefore the sensitivity of the test (35). The original

reports of PGS from the early and mid-1990s used polar body biopsies in conjunction with FISH to

screen the sex chromosomes and one, two or three autosomes (32,36). Polar body biopsies have the

advantage of not interfering with fertilization and subsequent embryo development, although the

safety of this technique has not been rigorously evaluated (37). Conversely, polar body biopsies do

not assess the paternal genetic contribution or detect post-zygotic abnormalities and are therefore

associated with unacceptably low detection rates (6,34,35,37).

8

Because of these drawbacks, cleavage stage biopsies of one or two blastomeres combined with FISH

became the standard method of performing PGS up until the mid- to late-2000s (34,35). There were

numerous clinical studies on this method of aneuploidy screening including a Cochrane review (38)

and subsequent meta-analysis (39). The Cochrane review included nine studies that used FISH to

screen between five and nine chromosomes. The LBR was found to be significantly lower in the PGS

arm whereas no difference was found in the rates of multiple pregnancies or early pregnancy losses

(38). The subsequent meta-analysis found strikingly similar results (39).

There are multiple explanations as to why PGS using blastomere biopsies and FISH is associated

with poorer clinical outcomes. First, embryo biopsy at the blastomere stage has been associated with

significantly lower implantation rates compared with either unbiopsied embryos or those biopsied at

the blastocyst stage (40). As such, there is justifiable concern that damage from the biopsy itself

more than offsets any potential gains from knowledge of ploidy status (40). Second, because FISH

screens only a limited number of chromosomes many of the embryos transferred after undergoing

PGS with FISH may in fact still be aneuploid (38). FISH can also lead to false positive results due to

scoring errors of FISH signals which can result in the non-transfer of euploid embryos (39,41). Third,

human cleavage stage embryos are known to have high rates of mosaicism – when two or more cells

from within the same embryo have a different chromosomal complements (42). It follows, therefore,

that PGS results of one or two cells may not, in fact, represent the underlying genetics of the majority

of that embryo. Furthermore, it is felt that mosaic embryos may be able to self-correct and become

euploid between the cleavage and blastocyst stages (39,42,43). Given these facts, it is easy to

understand, in retrospect, why this technology did not herald improvements in embryo selection nor,

ultimately, in IVF outcomes.

2.1.2 Trophectoderm Biopsy

The currently preferred method for obtaining genetic material for PGS is via trophectoderm biopsy at

the blastocyst stage of embryo development. This technique was first described on human embryos in

1990 (44), however the earliest reports of live births after trophectoderm biopsy did not occur until

approximately fifteen years later (45). The shift to trophectoderm biopsy was buoyed by parallel

advances in IVF laboratory techniques. Improvements in embryo culture techniques allowed embryo

culture to the blastocyst stage to become commonplace and the shift from slow freezing to embryo

vitrification allowed for embryos to be dependably cryopreserved while awaiting PGS results without

compromising cycle outcomes (35). Trophectoderm biopsy, unlike blastomere biopsy, has been

shown to preserve the reproductive potential of the embryo (37,40). Furthermore, it provides more

9

genetic material for analysis – five to ten cells compared to one or two – thus reducing the risk of

inaccurate results, particularly relating to mosaicism (35). Finally, the natural embryo selection

process that occurs between the cleavage and blastocysts stages of development results in a more

efficient process as only those embryos that had the intrinsic ability to reach the blastocyst stage are

biopsied (35).

2.1.3 Comprehensive Chromosome Screening

Just as trophectoderm biopsy has supplanted blastomere biopsy as a technique for obtaining

embryonic genetic material, comprehensive chromosome screening (CCS) has supplanted FISH as a

technique for analyzing that material. CCS involves one of a number of techniques all of which share

the same central feature – all chromosomes are evaluated (35,46).

2.1.3.1 Array Comparative Genomic Hybridization

Array comparative genomic hybridization (aCGH) is a technique whereby test and control DNA

samples are amplified and labeled with differently coloured fluorescent probes. The two samples are

then mixed together and allowed to competitively hybridize onto a DNA microarray. The ratios of

hybridization signal intensities are then analyzed to determine copy number (41,47,48). Early studies

of PGS using aCGH relied on blastomeres for source material (49), however the technique was

subsequently validated on trophectoderm samples (50,51). Despite its obvious and substantial

advances over FISH technology, it is limited in its inability to detect polyploidies, uniparental

disomy, structural chromosomal defects or abnormalities in areas of the genome not covered by the

microarray (35,41,48,52).

2.1.3.2 Single Nucleotide Polymorphism Analysis

Single nucleotide polymorphisms (SNPs) are biallelic single base pair variants that occur throughout

the genome. SNP-based CCS microarrays use hundreds of thousands, or millions, of SNPs to

determine copy number variations including aneuploidies and unbalanced translocations (41). This

can be done through qualitatively, quantitatively, or through a combination of both techniques (52).

SNP microarrays cannot consistently detect polyploidy and may not produce a result in the case of

parental consanguinity (48) but unlike aCGH they can detect balanced translocations and uniparental

disomy. Additionally, SNP microarrays have the advantage of being able to detect single gene

defects and screen for aneuploidy simultaneously (41,52,53).

10

2.1.3.3 Real-Time Quantitative PCR

Aneuploidy screening using real-time quantitative PCR (qPCR) followed both aCGH and SNP

microarrays in the chronology of CCS technologies, with the first reports published in 2012 (54).

qPCR involves the amplification of multiple loci spanning all 24 chromosomes followed by rapid

quantification of PCR products to detect aneuploidies (54–56). Because the sample is amplified

directly rather than relying on whole genome amplification, qPCR-based CCS cannot be performed

on single blastomeres and requires a trophectoderm sample (57). qPCR can produce results in only a

few hours, eliminating the need to vitrify blastocysts while awaiting results; this advantage, however,

is somewhat offset by the fact that the analysis of multiple biopsy samples in parallel is very labour

intensive or requires a significant investment in additional equipment (54,57). qPCR can detect

polyploidies but is unable to detect structural chromosomal defects or uniparental disomy (48).

2.1.3.4 Next Generation Sequencing

The newest technology being applied to CCS is next generation sequencing (NGS). NGS involves

the fragmentation and subsequent amplification of DNA, resulting in extremely large numbers of

short DNA sequences. These fragments are sequenced in parallel until a sufficient sequencing depth

is reached. They are then counted and compared with a reference human genome (57,58). This

technology can also be used to detect single gene mutations, polyploidies and uniparental disomy

(48) and, depending on the platform used, large duplications and deletions (48). It is significantly less

expensive than some of the other techniques described and is quickly emerging as the gold standard

in the field of CCS (58,59).

2.2 The Efficacy of Comprehensive Chromosome Screening

There are at least four published systematic reviews that have attempted to summarize the

observational and trial data on PGS using CCS technology (10,60–62). Dahdouh et al. has published

one systematic review and one meta-analysis on three randomized controlled trials (RCTs) that all

included good prognosis patients and used either aCGH or qPCR after trophectoderm biopsy. They

found increased implantation and ongoing pregnancy rates with the transfer of one or two euploid

embryos as compared to an unscreened embryo transfer (10,60). Unlike the RCTs, the eight cohort

studies tended to include poor prognosis patients and included polar body, blastomere and

trophectoderm biopsy samples. A meta-analysis of these studies showed a statistically significant

improvement in implantation rates compared to usual care despite a high degree of heterogeneity

11

(10). Lee et al. reviewed the same three RCTs in addition to five cohort studies and eleven case

series and also found a consistent improvement in implantation rates with the use of CCS (61).

Finally, Chen et al. performed a meta-analysis of four RCTs (three of the four being the same as

above plus one published abstract) and seven cohort studies (62). They also found higher

implantation rates in the RCTs and the observational studies; however, only the cohort studies, not

the RCTs, showed statistically significant increases in clinical and ongoing pregnancy rates and

lower EPL rates. Conversely, the single RCT to report LBR was able to show a benefit with CCS

whereas the pooled estimate from three cohort studies did not (62). Many of the analyses suffered

from very high levels of heterogeneity.

It is worthwhile to note that none of the above studies measured or reported on what is quickly

emerging as the gold standard of IVF outcomes – the cumulative live birth rate (CLBR) per cycle

(63). This statistic captures the outcomes of sequential embryo transfers resulting from one cycle of

COH and OPU, regardless of the number of transfers and whether or not those transfers were of fresh

or frozen embryos. Although the precise methodological details of measuring the CLBR are still

being debated amongst experts in the field (63), this statistic is particularly appropriate for studies of

PGS. First, eSET is more likely to be performed with euploid versus unscreened embryos making it

challenging, and perhaps altogether inappropriate, to interpret implantation and EPL rates after the

first transfer. Second, PGS by its very nature reduces the size of the transferrable cohort of embryos,

the effects of which may not be evident unless all transfers from a given OPU are fully compared.

In summary, while CCS at the blastocyst stage has been shown to improve implantation rates per

embryo transferred, its impact on EPL and LBR have yet to be confirmed in well-controlled large

scale RCTs. High quality randomized trial data comparing CCS using NGS to no PGS has yet to be

published and the impact of PGS on CLBRs has not been established in the literature.

2.3 The Cost-Effectiveness of PGS

2.3.1 CEAs of Older Methods of PGS

There is a single CEA of PGS using blastomere biopsy and FISH technology. This study looked at

women aged 38 – 40 and showed a higher average cost per euploid infant with PGS (ACER =

$118,713) compared to IVF alone (ACER = $68,026) from a societal perspective (64). Given what is

now known about the differences between PGS using blastomeres and FISH and trophectoderm and

CCS (see Section 2.1.1), this study is no longer applicable.

12

2.3.2 CEAs of PGS for Recurrent Pregnancy Loss

One full length publication and one published abstract have looked at the cost-effectiveness of PGS

using CCS for patients with recurrent pregnancy loss (RPL) (65,66). These studies, while worth

reviewing, are not directly applicable to the general infertility population because those suffering

from RPL often do not have concurrent infertility.

The published abstract by Resetkova et al. found a lower average cost per live birth with IVF and

PGS (ACER = $19,416) compared to IVF alone (ACER = $23,184) from a societal perspective (65).

The study by Murugappan et al. compared IVF with PGS (using a combination of blastomere and

trophectoderm biopsies plus aCGH) to expectant management (i.e. spontaneous conception) in

patients with RPL. They found a much higher average cost per live birth in the IVF and PGS arm

(ACER = $45,300) compared with expectant management (ACER = $418). This resulted in an ICER

of $71,906 per additional live birth (66). Furthermore, although this CEA did not model time to

conception, a retrospective cohort study by the same authors found that the time to conception was

significantly shorter in the expectant management group compared with the group that chose to

pursue PGS for RPL (67). Therefore, according to this pair of studies, IVF with PGS for patients

with RPL but without infertility is more expensive, less effective and associated with a longer

duration of childlessness compared to expectant management.

2.3.3 CEAs of PGS using CCS

To date, there are only five published abstracts and one full-length paper that attempt to address the

cost-effectiveness of CCS, with mixed findings. This small body of literature highlights the need for

the current study. These publications are summarized below.

Methods Key Findings Limitations

Abstracts

Resetkova

et al., 2014

(68)

Modelled IVF/PGS using

aCGH compared to IVF

alone to achieve a live

birth. Patient perspective.

Maximum of three

pregnancy attempts.

Average cost per live birth

was higher with PGS vs.

IVF alone. PGS was

approximately $2500

more expensive that IVF

alone and increased the

success rate from 64% to

95%.

Number of embryos per

transfer not specified.

Maximum of three

transfers were modelled.

Unclear if productivity

costs included.

Patounakis

et al., 2014

(69)

Modelled the cost-

effectiveness of euploid

DET vs. unscreened DET,

euploid SET vs.

unscreened SET and

Scenarios in which PGS

required the freezing of all

embryos followed by

frozen embryo transfers

were not cost-effective.

Perspective not specified.

Definition of “cost-

effective” not specified.

Pregnancy and live birth

rates used were obtained

13

euploid SET vs.

unscreened DET. Fresh

vs. frozen transfer with

PGS was also modelled.

When the costs of

multiple births were

included, PGS was most

cost-effective when used

to promote single embryo

transfer.

from a few small studies

that are not generalizable

to the general infertility

population.

Hodes-

Wertz et al.,

2015 (70)

Retrospectively compared

the costs of IVF/PGS and

IVF cycles from 2011 to

2013.

The average cost per

delivery was equivalent

for women < 35 years old,

lower with IVF/PGS for

women aged 35 – 39 and

40 – 42 but much higher

for women > 42 years old.

The multiple gestation rate

was substantially lower

with IVF/PGS.

Number of embryos per

transfer not specified.

Unclear if any patient

factors, such as

diagnosis, reason for PGS

or previous treatment

outcomes, were

controlled for.

Neal et al.,

2016 (71)

Modelled costs of

sequential unscreened

SETs from one OPU until

either live birth was

achieved or no embryos

remained. Model was

applied and compared to a

retrospective cohort of

nearly 12,000 IVF/PGS

cycles.

Average cost per delivery

or exhaustion of all

embryos was

approximately $800

higher in the IVF vs. PGS

group. There were two

and a half times more

transfers in the IVF alone

vs. PGS group.

Estimates from

retrospective cohort may

not be generalizable as

included only those that

actually underwent

IVF/PGS.

Did not include dropout

in the model.

Salem et al.,

2016 (72)

Designed a model to

compare IVF/PGS with

IVF alone to achieve live

birth in women >= 39

years old. Tested false

positive PGS rates from

4% – 16%.

IVF/PGS resulted in lower

live birth rates than IVF

alone for all age groups

studied. The average cost

per live birth was higher

with IVF/PGS compared

to IVF alone.

Perspective not specified.

Maximum number of

transfer attempts unclear.

Full Length Publications

Scriven,

2016 (73)

Modelled sequential

euploid vs. unscreened

SETs up to a maximum of

10 to achieve an ongoing

pregnancy or exhaust

embryo cohort. Both fresh

and freeze-all strategies

tested. Societal

perspective.

IVF/PGS did not improve

the cumulative live birth

rate compared with IVF

alone but did decrease the

number of transfers and

early pregnancy losses.

IVF/PGS resulted in fewer

live births per cycle as its

positive predictive value

decreased.

Included productivity

costs were not

comprehensive. PGS

outcomes estimated from

two small studies (51,74).

Dropout rate assumed to

be zero.

There is significant heterogeneity in both the methodology and findings of these cost-effectiveness

studies. The perspective of the analysis is an important consideration as the majority of the above

studies included only direct ART costs and it is unclear from the abstracts alone if medication costs

14

were included in these estimates. None of these studies included the costs of comprehensive

productivity losses that account for both partners’ time off work due to IVF or its complications. The

number of transfer attempts and the number of embryos per transfer are critical considerations when

evaluating any study of PGS effectiveness or cost-effectiveness as these variables will affect both the

clinical outcomes and the direct costs. Comparing the cost-effectiveness of, for example, three

sequential single unscreened embryo transfers with three sequential single euploid embryo transfers

until a live birth is achieved is not a clinically meaningful comparison because not all embryos are

euploid and therefore the large majority patients with three unscreened embryos available for transfer

will have fewer than three euploid embryos available after PGS with which to attempt to achieve a

pregnancy. With respect to the number of embryos transferred at time, studies looking at the cost-

effectiveness of transferring two euploid embryos at a time are difficult to apply to today’s clinical

practice given the recommendations to always transfer a single embryo at a time when transferring a

euploid embryo (17).

As discussed earlier, studies of PGS effectiveness should ideally examine the CLBR from one cycle

of COH and OPU, at least until the first live birth is achieved or treatment is discontinued by the

patients. Thus far, the published trials of PGS effectiveness have compared only the first transfer in

each group (10,62). Moreover, these trials have included only subsets of either good or poor

prognosis patients and are not representative of the general infertility population. Cost-effectiveness

findings based on clinical inputs from these studies are therefore only applicable to those specific

populations. Finally, none of the above studies modelled the clinical reality that some patients

prematurely discontinue treatment before achieving either a live birth or exhausting their embryo

cohort – either due to financial or psychological reasons.

Given the rapidly expanding use of PGS technology in everyday infertility practice and the scarcity

of publications on its cost-effectiveness, it is not surprising that experts in the field have been calling

for a more robust examination of this topic (61,75,76). The current study addresses many of the

limitations of the studies highlighted above and aims to considerably add to the body of literature on

this subject.

2.4 Study Objective

The objective of this study is to examine the cost-effectiveness of IVF/PGS using CCS technology

compared with IVF alone to achieve a live birth in a general infertility population in Ontario from a

societal perspective. The following female age groups are studied separately: less than 35 years, 35 –

15

37 years, 38 – 40 years, 41 – 42 years and greater than 42 years of age. The study will also examine

the cost-effectiveness of IVF/PGS from three alternative perspectives (direct health care perspective,

hypothetical single-payer government insurer perspective, and patient self-pay perspective). To

achieve this objective, a novel TreeAge Pro model to study CLBRs after IVF will be designed and

constructed.

16

3 Methods

3.1 Study Design

A decision analytic model containing multiple Markov model clones was created using TreeAge Pro

2016 to compare IVF/PGS to IVF alone. The time horizon was one complete IVF cycle, defined as

one cycle of COH and OPU followed by the sequential transfer of all usable embryos created as a

result of that fresh cycle until one of the primary end points was reached. These end points were live

birth, premature treatment discontinuation by the patient or exhaustion of all transferrable embryos.

For the base case analysis costs were calculated from a societal perspective. All embryo transfers

were modelled as single embryo transfers at the blastocyst stage.

This paper follows the Consolidated Health Economic Evaluation Reporting Standards (CHEERS)

recommendations as set out by the ISPOR Health Economic Evaluation Publication Guidelines Good

Reporting Practices Task Force (77).

3.2 Model Design & Structure

The decision tree begins with a decision node which leads to the two competing strategies: IVF/PGS

and IVF alone.

The IVF alone arm leads to a Markov model with three states: embryo transfer, drop out and live

birth. The entire cohort starts out in embryo transfer as this model assumes all subjects have at least

one unscreened blastocyst. The latter two states are absorbing states. Live birth refers to the clinical

scenario where the treatment ends successfully in a live birth. Drop out refers to the clinical scenario

where patients discontinue treatment despite not having achieved a live birth nor having exhausted

transferring all their embryos created from one round of COH and OPU. The embryo transfer state

branches into a subtree with the options of pregnant or not pregnant. The not pregnant arm branches

again into embryo transfer and drop out, both of which lead back to their corresponding states. The

pregnant arm branches into live birth and pregnancy loss. The former branch leads back to its

corresponding state whereas the latter leads to the same two options as not pregnant: embryo transfer

and drop out. This Markov model is shown diagrammatically in Figure 8.2.

17

The IVF/PGS arm leads to a series of nine identical Markov clones as well as a terminal branch

representing the possibility of having no euploid embryos available to transfer. Each Markov clone is

identical in structure to the one described above and shown in Figure 8.2.

The model was designed such that one cycle length equals one embryo transfer. A counter variable

entitled “Total_Blasts” was created and set to equal the total number of unscreened blastocysts

available. The IVF alone arm was made to terminate when the stage counter equaled the value of the

counter variable. The probability of routing towards one of the ten possible branches within the

IVF/PGS arm was determined by two variables: the total number of unscreened blastocysts (i.e. the

“Total_Blasts” variable) and the age category (< 35, 35 – 37, 38 – 40, 41 – 42 and > 42) being

studied. For every age category and starting number of blastocysts, the relative probabilities of

having a specific number of euploid blastocysts was calculated. For example, if the unscreened

embryo cohort size was three, the chances of having zero, one, two or three euploid embryos was

estimated for each of the age categories being studied. These expected euploidy rates were obtained

from the published literature (8). These values were then used to determine the probability of moving

from the initial IVF/PGS node to a specific Markov clone or, in the case of no euploid blastocysts

available, the no euploid terminal branch. For example, for women aged 35 – 37 with 2 unscreened

blastocysts, approximately 12.6% of the cohort would move to the no euploid node while 45.8%

would move to the 1 euploid Markov clone and the remaining 41.6% to the 2 euploid Markov clone.

Each of the nine Markov clones has a termination condition equal to the number of euploid blasts

represented by that clone. For example, the 1 euploid model terminates after one iteration, which

equates to one embryo transfer, while the 9 euploid one terminates after nine iterations, or embryo

transfers.

A simplified version of the IVF/PGS arm of the decision tree including the embedded Markov model

is shown in Figure 8.3. The full decision tree is shown in Figure 8.4.

3.3 Clinical Parameters

All three Markov states – embryo transfer, drop out and live birth – have initial and incremental

effectiveness values of zero. (It is important to note here and throughout Chapter 3 that the meaning

of the term incremental does not refer to the difference in effectiveness between two interventions or

two costs (i.e. the E or the C of the ICER) but rather the effectiveness or costs associated with

spending a Markov cycle in a given Markov state in TreeAge Pro.) The live birth state has a final

18

effectiveness value 1. The embryo transfer and drop out states have final effectiveness values of

zero.

3.3.1 CARTR Plus / BORN Ontario

The Canadian Assisted Reproductive Technologies Register (CARTR) is Canada’s national assisted

reproductive technology database (78). It is run by the Better Outcomes Registry and Network

(BORN) of Ontario – a provincial pregnancy, birth and childhood database. The original intention of

this study was to use Ontario data from 2013 obtained from CARTR Plus / BORN to inform our

model inputs with respect to pregnancy and live birth rates after IVF/PGS and IVF alone. This was

the first in which these two databases were merged therefore allowing the linkage of IVF cycles with

more detailed birth outcomes. Data from more recent years were not available at the time of our data

request.

In total, information from 8608 cycles, both fresh and frozen, was obtained from CARTR Plus /

BORN Ontario. Donor oocyte and gestational carrier cycles were excluded by the data analysts at

BORN Ontario as they were not representative of the general IVF population. We then further

excluded frozen oocyte IVF cycles (26 cases), PGD cycles (3 cases) and embryo or oocyte banking

cycles (29 cases) for the same reason. Cycles with no pregnancy or birth outcome data (19 cases)

were excluded for lack of data. Ectopic pregnancies (44 cases) and stillbirths (10 cases) were

categorized as pregnancy losses. Pregnancy losses were defined as clinical pregnancy losses and

therefore biochemical pregnancies (524 cases) were categorized as failed embryo transfers. There

were cases where the result of the embryo transfer was a clinical intrauterine pregnancy but the

ultimate pregnancy outcome was coded either as unknown (61 cases), missing (29 cases) or left

blank (42 cases). For these, birth outcome was categorized as a live birth if there was a fetal heart

beat present and as a pregnancy loss otherwise.

It became clear after further analysis of the data that there were significant limitations with respect to

how key information, particularly regarding PGS, was collected and/or coded. While CARTR Plus /

BORN Ontario data was used in this study whenever possible, many of the clinical parameters were

ultimately obtained elsewhere. CARTR Plus / BORN Ontario data was used to obtain: the average

dose of FSH for women undergoing either IVF alone or IVF/PGS, the percentage of non-PGS cycles

that used ICSI as the method of fertilization, the percentage of pregnancy losses attributable to

ectopic pregnancies after IVF, and the distribution of modes of delivery (spontaneous vaginal

19

delivery, operative vaginal delivery and Caesarean section) after IVF. A more detailed discussion of

CARTR Plus / BORN Ontario can be found in Chapter 6.

3.3.2 Pregnancy & Pregnancy Loss Rates

Pregnancy and pregnancy loss rates were obtained from the Society of Assisted Reproductive

Technology’s (SART) publicly available online National Summary Report for 2014 (79). Aggregate

data from the “Primary Outcome per Intended Retrieval” tab for patients undergoing conventional

stimulation IVF cycles with patients’ own oocytes were used. The data were filtered to exclude

gestational carrier and frozen oocyte cycles and to include only single embryo transfer cycles on day

5 or 6. PGS/PGS cycles were compared to non-PGS/PGD cycles to obtain model inputs. Table 7.1

shows the clinical input parameters used in the model.

3.3.3 Treatment Discontinuation

It is well known that many patients discontinue infertility treatment before achieving a live birth due

to non-medical reasons such as psychological distress, relationship issues or financial concerns (80–

82). Premature treatment discontinuation or withdrawal is often referred to in the literature as

dropping out. While acknowledging the negative connotations associated with this term (83), the

term dropout will nonetheless be used here for concision and consistency with previous publications.

The dropout rate has traditionally been defined as the proportion of eligible patients not returning for

a subsequent fresh IVF cycle (comprised of COH, OPU and embryo transfer(s)) after the failure to

achieve a live birth in a previous cycle. Studies on dropout rates have consistently found rates

between 15% and 30% (84–90). Premature dropout is, in fact, included in many other decision

models of IVF (for examples see (91–93)). However, nearly all the studies on dropout rates have

focused on this between-cycle dropout rate and, as a result, very little is known about within-cycle

dropout rates – the probability of discontinuing treatment prior to transferring all cryopreserved

embryos from a single cycle of COH and OPU. The data obtained from BORN/CARTR could not be

used to calculate either between-cycle or within-cycle dropout rates for Ontario. Therefore, a

thorough review of the literature was undertaken in an attempt to estimate within-cycle dropout rates

for use in our model.

There are many studies, both quantitative and qualitative, that have examined patients’ decision

making with regards to the disposition of surplus embryos (94–99). These studies included varying

degrees of demographic details, including parity and family status; however, none of these studies

20

provided enough information to determine what percentage of infertile patients with embryos

remaining choose not to use them for another attempt at a pregnancy. A study by Domar et al. looked

at couples who had insurance coverage for up to three cycles of IVF but discontinued treatment

before their third cycle despite not achieving a pregnancy (83). They found that one year after

starting their first cycle 30% of these patients had frozen embryos remaining and that only half of

them returned to use their embryos within the subsequent three years. Therefore, if the between-cycle

dropout rate is approximately 15-30% and the within-cycle dropout rate is approximately 15-30% of

that, it follows that the within-cycle dropout rate can be estimated at 5.60% (range 2.25% to 9.00%).

Additionally, there is some evidence to support the fact that the dropout rate after a pregnancy loss is

higher than after a failed transfer. A recent abstract reported a hazard ratio of 1.18 (90). Therefore,

the within-cycle dropout rate after a pregnancy loss was estimated at 6.61% (range 2.66% to

10.62%).

3.4 Cost Parameters – Societal Perspective

The costs included in the present model, from a societal perspective, encompassed direct health care

costs related to IVF and PGS, health care costs related to prenatal care, and productivity costs related

to time off work for both patients and their partners. Costs are provided in 2015 Canadian dollars. A

discount rate of 0% was applied because the time horizon of the model was short and because it was

modelled in embryo transfer attempts rather than absolute units of time.

3.4.1 Assisted Reproductive Technologies & Medication Costs

The costs of infertility treatments were drawn from fee schedules of Ontario fertility clinics. These

were obtained from publicly available websites as well as through personal communications with

fertility clinic staff wherever possible. Online fee schedules that had not been updated in the previous

two years were excluded as they were assumed to out of date and inaccurate. A list of these costs can

be found in Table 7.1.

Medication costs associated with fertility treatments were calculated in the following manner. The

average dose of FSH used per fresh cycle was obtained from BORN/CARTR. The average FSH dose

was 2903.58 IU for those undergoing IVF alone and 3730.52 IU for those undergoing IVF/PGS.

These values were used regardless of age cohort. The average cost per unit FSH was calculated by

averaging the costs of the two most common FSH formulations used locally: Gonal-f® (EMD

21

Serono) ($350 for 300 IU) and Puregon® (Merck) ($360 for 300 IU). It was assumed that all cycles

were GnRH antagonist cycles with ten days of Estrace® (Acerus) priming (4mg per day) followed by

FSH alone. The cost of Estrace® was set at $0.53 per 2mg. It was assumed that each patient received

four days of a GnRH antagonist at $125.00 per day and one dose of Ovidrel® (EMD Serono) trigger

at $85.00. Finally, it was assumed that each patient took 200mg of Prometrium® (Merck) twice daily

for fourteen days for luteal support. The cost of Prometrium® was set at $1.67 per 100mg. This

protocol and these costs were based on expert opinion and are consistent with local practice.

Frozen embryo transfer (FET) cycles were assumed to be medicated with 4mg of Estrace® twice

daily for fourteen days followed by two weeks of 2mg of Estrace® twice daily and 200mg of

Prometrium® three times daily. Patients who became pregnant after a fresh embryo transfer were

assumed to continue their luteal support until 7 weeks of gestation. Those who became pregnant after

an FET were assumed to continue their luteal support until ten weeks of gestation. These protocols

were based on expert opinion and are consistent with local practice.

It is acknowledged that some IVF cycles, and PGS cycles in particular, may be triggered with a

GnRH agonist rather than hCG and followed up with blood work the next day. Similarly, many other

types of FET cycles exist as do other options for luteal phase support. There is also debate regarding

the optimal duration of luteal phase support. These alternatives were not explicitly modelled in this

study; however, a sensitivity analysis was performed that varied the cost of medications associated

with both fresh and FET cycles (see Section 3.7). In order to estimate a range of values for the

purposes of sensitivity analyses, medication costs were multiplied by 2/3 (minimum) and 4/3

(maximum). The final medication costs used in the model can be found in Table 7.2.

3.4.2 Productivity Costs

Estimates of amount of work missed due to IVF and any resultant pregnancies were estimated based

on a combination of local practice, expert opinion and, wherever available, published resources

(92,100–106). Women undergoing COH and OPU were estimated to miss a total of fifteen hours of

work: six hours over the course of COH plus one eight hours work day for the OPU. Similarly, it was

estimated that a fresh ET would result in four hours of missed work, an FET in five hours of missed

work, routine first trimester care in six hours of missed work and routine second and third trimester

care in a combined fourteen hours of missed work. It was assumed for this study that women

undergoing IVF had partners. It was estimated that these partners would miss ten hours of work

22

during the course of their partner’s COH and OPU and four hours of work for both fresh and frozen

embryo transfers. These estimates were in keeping with other published estimates (92,101–105)

although large differences in study designs, patient characteristics, changes in practice patterns over

time and cultural factors makes direct comparisons and extrapolations challenging. No published

estimates were available from a Canadian setting.

Productivity losses associated with pregnancy loss were obtained from a published economic

evaluation done alongside a British RCT examining various methods of EPL management (106).

Pregnancy losses managed expectantly, medically and surgically were therefore allocated 9.37, 10.63

and 9.86 days off work, respectively (106). Productivity losses associated with ectopic pregnancies

were assumed to be fifteen days, or approximately 1.5 times the number of days associated with EPL,

regardless of whether the ectopic pregnancy was treated medically or surgically.

For the purposes of calculating productivity costs for the present model, all workdays were assumed

to be eight hours in length. Productivity costs associated with missed work for partners of women

undergoing prenatal care, pregnancy complications or labour and delivery were not included.

Estimates of work absences were converted to costs using the average hourly wage rate for females

($25.77/hr.) and males ($29.85/hr.) ages 25 to 54 in Ontario for 2015 (107). Although clearly not

always the case, it was assumed for the purposes of wage rate calculations in this study that patients

were female and their partners were male. In order to estimate a range of values for the purposes of

sensitivity analyses, time off work estimates were multiplied by 2/3 (minimum) and 4/3 (maximum).

The final productivity costs included in the model can be found in Table 7.2.

3.4.3 Hospital and Physician Costs

The Ontario Case Costing Initiative (OCCI) is an initiative of the Ministry of Health and Long-Term

Care of Ontario, Canada that collects cost data for acute inpatient, day surgery and ambulatory care

cases. This initiative has an online Costing Analysis Tool (CAT) that can be used to obtain cost

estimates based on fiscal year, hospital, principal procedure, most responsible diagnosis, age group

and Case Mix Groups (categories developed by the Canadian Institute for Health Information) (108).

The CAT from the most recently available fiscal year (2010/2011) was used to obtain hospital costs

for the following model inputs: dilation and curettage, surgical management of ectopic pregnancy,

medical management of ectopic pregnancy, spontaneous vaginal delivery, operative vaginal delivery,

23

Caesarean section and neonatal costs. All costs were adjusted to 2015 Canadian dollars using the

consumer price index, Health Care only, Ontario only (109).

The Schedule of Benefits, published by the Ontario Ministry of Health and Long Term Care

(MOHLTC), details physician fees for all physician services covered by Ontario’s provincial health

plan (110). This was used to obtain physician costs for the following model inputs: prenatal visits

(assumed 12 throughout pregnancy), first trimester ultrasound, anatomy ultrasound, medical

management of pregnancy loss, dilation and curettage, medical management of ectopic pregnancy,

surgical management of ectopic pregnancy, spontaneous vaginal delivery, operative vaginal delivery

and Caesarean section. For ultrasounds both technical and professional fees were included. Where

applicable, anesthesia fee codes were also included and based on estimates of procedure lengths.

Special visit premiums, including after-hours premiums, were not included.

The cost of routine prenatal screening was obtained based on raw data used in a 2014 publication by

Okun et al. and obtained via personal communication with the author (20). Hospital and physician

costs can be found in Table 7.2.

3.4.4 Calculating Costs within the Decision Tree

The following sections describe how the individual costs described in Table 7.2 were combined and

distributed to calculate the costs associated with different health states, and the movement between

those states, of the Markov models embedded within our decision tree.

3.4.4.1 Initial Costs: Health States

Initial costs are those costs associated with starting out in a particular Markov health state. In the

present model, 100% of subjects started in the embryo transfer state and no subjects started in either

the dropout or live birth states (see Section 3.2). The initial costs associated with the embryo transfer

state can be thought of as the sum of all costs up to and including the first embryo transfer. The initial

costs for the drop out and live birth states were set to zero because no subjects started out in those

states.

The initial cost for the embryo transfer state of the Markov model within the IVF alone arm of the

decision tree was calculated by summing the following individual costs:

- IVF cycle

- IVF cycle medications

24

- COH and OPU productivity

- COH and OPU partner productivity

- fresh embryo transfer productivity

- fresh embryo transfer partner productivity.

It also included the cost of ICSI multiplied by the proportion of cycles that used ICSI, which

according to CARTR Plus / BORN Ontario data, was 66% of all IVF cycles (111).

The initial cost for the embryo transfer state of the Markov model and its clones within the IVF/PGS

arm of the decision tree was calculated by summing the following individual costs:

- IVF cycle

- ICSI

- embryo biopsy and PGS

- embryo cryopreservation

- IVF cycle medications

- COH and OPU productivity

- COH and OPU partner productivity

- FET cycle

- FET medications

- FET productivity

- FET partner productivity.

The key differences between these costs and the initial costs for the embryo transfer state of IVF

alone were: the use of ICSI 100% of the time, the need for embryo biopsy, PGS and

cryopreservation, and the fact that the first embryo transfer was a frozen embryo transfer rather than

a fresh one.

The costs associated with the No Euploid branch of the IVF/PGS arm of the decision tree was

calculated by summing the following individual costs:

- IVF cycle

- ICSI

- embryo biopsy and PGS

- embryo cryopreservation

- IVF cycle medications

- COH and OPU productivity

25

- COH and OPU partner productivity.

The key difference between these costs and the initial costs of the embryo transfer state of the

Markov model within the IVF/PGS arm of the decision tree was the absence of costs associated with

a frozen embryo transfer.

3.4.4.2 Incremental Costs: Health States

As previously mentioned, TreeAge Pro uses the term incremental costs to refer to costs accrued by

spending any cycle other than the initial cycle in a given Markov state. While this terminology may

be confusing to many it is used here nonetheless for consistency with the TreeAge Pro software and

those familiar with it.

For every subsequent Markov cycle spent in the embryo transfer state of the Markov model within

the IVF alone arm accrued costs were calculated as the sum of:

- FET cycle

- FET cycle medications

- FET productivity

- FET partner productivity.

The cost of cryopreservation was included only if it was the second embryo transfer. The cost of an

additional year of embryo storage was included if it was the sixth embryo transfer. This assumed that

those having five or fewer transfers will complete them within one year and those having six or more

will complete them over two years.

The incremental cost associated with the embryo transfer state of the Markov model and its clones

within the IVF/PGS arm of the decision tree was calculated as the sum of:

- FET cycle

- FET cycle medications

- FET productivity

- FET partner productivity.

There were no incremental costs associated with spending time in the other states because they were

both absorbing states.

3.4.4.3 Markov Model Transition Costs

Transition costs are the costs accrued when moving from one Markov state or subtree to another.

26

The costs associated with pregnancy loss were captured when subjects transitioned from pregnant to

pregnancy loss within the subtree of the embryo transfer health state. The overall cost of pregnancy

loss was calculated as the sum of:

- prenatal consult

- early ultrasound

- first trimester luteal support medications (calculated as a weighted average of luteal support

medication costs from both fresh and frozen embryo transfer cycles in the IVF alone arm)

- first trimester care productivity.

The probabilities of the five treatment options (early pregnancy loss treated expectantly, early

pregnancy loss treated medically, early pregnancy loss treated surgically, ectopic pregnancy treated

medically, ectopic pregnancy treated surgically) were multiplied by their associated hospital,

surgeon, anesthesia and productivity costs and then summed and added to the above costs.

Based on data from CARTR Plus / BORN Ontario, 7.9% of all pregnancy losses were due to ectopic

pregnancies (111). Data was not available with respect to what percentage of ectopic pregnancies

were treated medically versus surgically. Two US studies found that approximately one third of

ectopic pregnancies were treated medically whereas two thirds were treated surgically (112,113). A

third study, however, found the opposite – approximately one third were treated surgically and two

thirds medically (114). For the current model, we assumed that half of the ectopic pregnancies, or

3.9% of the pregnancy losses, were treated surgically and the other half medically. With respect to

the management of intrauterine pregnancy losses, the data is equally sparse. Data were not available

to inform the distribution of expectant, medical and surgical management options for EPL in cases

where patients were informed of all three options and subsequently allowed to choose the one that

was most appropriate for them. Therefore, for this study we assumed that one third of all intrauterine

pregnancy losses were treated with each of the three management options.

The costs associated with live birth were captured when subjects transitioned from pregnant to live

birth within the subtree of the embryo transfer health state. The cost was calculated as the sum of:

- prenatal consult

- early ultrasound

- anatomy ultrasound

- prenatal screening

- first trimester care productivity

27

- second and third trimester care productivity

- twelve prenatal visits

- neonatal care.

The cost of first trimester luteal support medication, calculated as a weighted average of luteal

support medication costs from both fresh and frozen embryo transfer cycles for the IVF alone arm

was also included. Finally, costs associated with delivery itself were included. The probability of

each mode of delivery (spontaneous, operative vaginal and Caesarean section) was multiplied by the

sum of costs related to that mode of delivery (hospital, surgeon and, where applicable, anesthesia).

These probabilities were obtained from CARTR Plus / BORN Ontario data (111) and were as

follows: spontaneous vaginal delivery (0.446), operative vaginal delivery (0.123) and Caesarean

section (0.431).

3.5 Model Assumptions

While decision models are extremely useful tools for examining the cost-effectiveness of various

interventions, they are necessarily associated with simplifications of the clinical reality being studied.

It is also frequently necessary to make some assumptions regarding how best to use available clinical

and cost data to inform input parameters. A number of these assumptions are detailed below and

discussed in greater detail in Chapter 6.

The model accounted for an increased dropout rate after EPL compared to after a failed embryo

transfer; however, the model assumed that the dropout rate was independent of the number of

previous embryo transfers and whether or not those transfers were of euploid or unscreened embryos.

There is some data that the between-cycle dropout rate depends on patient factors such as treatment

prognosis, age and treatment history (81,90,115). Comparable data on within-cycle dropout rates was

not available; therefore, static dropout rates were used.

Pregnancy loss and live birth rates were modelled independent of cycle number. In other words, the

chance of having a live birth with a given embryo transfer was assumed to be the same regardless of

whether it was a the first embryo transfer or not. This model also assumed, consistent with current

practice and supported by cohort studies, that the EPL and LBRs were not significantly different

between fresh and frozen embryo transfers (116).

The model structure was designed to accommodate no more than ten available blastocysts. This

number is much higher than many of the other models that have been designed to study PGS in

28

which the maximum number of pregnancy attempts was limited to two (64,66). Furthermore,

according to data from CARTR Plus / BORN Ontario (111), fewer than 5% of all IVF cycles in

Ontario in 2013 resulted in more than ten unscreened blastocysts available for transfer; therefore, this

model encompasses the vast majority of cases encountered in clinical practice.

In this study, we assumed that all surplus embryos successfully survived being cryopreserved and

subsequently rethawed. We also assumed that all pregnancy losses were either first trimester losses

or ectopic pregnancies, rather than stillbirths. According to data from CARTR Plus / BORN Ontario

(111) less than 0.2% of all embryo transfers resulted in stillbirths. This number was too small to

justify its inclusion in the model structure as this would have resulted in much more complex Markov

subtrees.

Finally, the positive predictive value of PGS was assumed to be 100%. In other words, it was

assumed that all embryos with aneuploid PGS results were truly aneuploid. In reality, the true test

performance of PGS is unknown, and although it is generally considered to be very accurate, no test

is perfect. There are numerous methodological challenges associated with determining the

performance characteristics of PGS, including identifying the appropriate gold standard for

comparison and obtaining genetic material from a blastocyst’s inner cell mass. In general, however,

the data does support a sensitivity and specificity of greater than 98% (48,117–120).

3.6 Alternative Perspectives

In addition to the societal perspective, the model was also examined from three alternative

perspectives:

1. Direct health care – this perspective includes all of the costs of the societal perspective with

the exception of productivity costs, which were excluded from this analysis. This perspective

includes costs born both by the government, for example, the cost of prenatal care, and by the

patients, for example, the cost of PGS or fertility medications.

2. Hypothetical MOHLTC – this perspective was designed to simulate an entirely hypothetical

scenario whereby the MOHLTC covers the cost of PGS, but not any of the other costs related

to fertility treatment, in addition to the other direct health care costs that they currently cover.

29

3. Patient self-pay – this perspective includes only the direct health care costs related to fertility

treatments and fertility medications. It does not included any productivity costs nor does it

include any of the costs associated with normal or abnormal pregnancies and their care.

Table 7.3 clearly outlines which costs are included in which perspectives.

3.7 Sensitivity Analyses

In order to explore the uncertainty around the estimates, both one-way and probabilistic sensitivity

analyses were performed. Sensitivity analysis involves altering model inputs to understand how they

impact model results (26). In a one way sensitivity analysis, the value of a single variable is altered

and its impact on the results examined (19,26). With probabilistic sensitivity analysis (PSA), many

variables are changed simultaneously. Each parameter is first assigned a distribution shape and range.

The model is then run repeatedly using Monte Carlo simulation with the point estimates for each

parameter for each run being chosen at random from within the assigned distribution for each

parameter. This allows the joint uncertainty across all parameters to be studied simultaneously

(19,26,121).

3.7.1 One-way Sensitivity Analysis

One-way sensitivity analyses were performed on the variables listed in Tables 7.1 and 7.2 in order to

characterize uncertainty around the estimates and identify key drivers of the model. The ranges used

for clinical and cost inputs can be found in Tables 7.1 and 7.2. Because the SART National Outcome

Summary does not provide ranges or standard deviation estimates, a range of +/- 15% was used. The

range of values for dropout rates (see Table 7.1) were determined from a review of the available

literature (see Section 3.3.3). The minimum and maximum values for fertility treatment costs were

based on the actual range of prices charged at different fertility clinics in Ontario (122). The range of

values for medication costs and productivity costs were determined based on local practice and

expert opinion. Costs obtained from the Ontario MOHLTC Schedule of Benefits, with the exception

of anesthesia-related costs which are partially time based, were not evaluated in one-way sensitivity

analyses as these costs were known precisely. The range of anesthesia costs were varied based on

expert opinion. Minimum and maximum values for hospital costs were obtained from the OCCI.

ICER tornado diagrams were used to summarize the results of the one-way sensitivity analyses.

30

3.7.2 Probabilistic Sensitivity Analysis

In addition to one-way sensitivity analyses, probabilistic sensitivity analysis (PSA) was performed.

All relevant variables were defined with distributions as shown in Tables 7.1 and 7.2. Wherever

possible, costs were defined using gamma distributions as per accepted practice (26). Probabilities

were defined with triangle distributions (rather than beta distributions) because information about

standard deviations was not available. One thousand iterations were performed for each PSA. Scatter

plots of the ICERs from each iteration were displayed on CE planes in order to summarize the results

of the PSAs. Additionally, PSA results were presented in a table detailing what percentage of

iterations resulted in ICERs in each of the four quadrants of the CE plane.

3.8 Ethics Approval

Ethics approval was obtained from the Sinai Health System research ethics board. Approval was also

obtained from the University of Toronto and BORN Ontario.

31

4 Results

Regardless of perspective taken, PGS became increasingly cost-effective – more effective and less

costly – as the number of embryos available and the age of the woman increased.

4.1 Societal Perspective

The point estimates for C, E and ICERs by age group and by the number of blastocysts available

can be found in Table 7.4.

4.1.1 Incremental Cost

When analyzed from the societal perspective, the model found that IVF alone was less costly than

IVF/PGS for women 42 years of age or younger, regardless of the number of blastocysts available.

For women less than 35 years old, the incremental cost of PGS ranged from $4,928 to $6,801. For

women aged 35 to 37 the results were similar with a range of $5,088 to $6,307. The incremental

costs began to decrease in women aged 38 to 40, particularly with larger blastocyst cohorts, with C

ranging from $3,268 to $5,840. This trend continued for women aged 41 to 42 but with an

appreciable widening of the range of values, with a C of $22 with ten blastocysts and $5,632 with

one. IVF/PGS remained more costly than IVF alone for women greater than 42 years of age with

one, two or three blastocysts (C $1,064 to $5,437) but became less costly once four or more

blastocysts available to biopsy. The cost savings ranged from $289 to $4,926.

4.1.2 Incremental Effect

The model found that IVF alone was associated with a higher probability of live birth compared to

IVF/PGS for women less than 35 years of age, regardless of the number of blastocysts available. For

women aged 35 to 37, the difference in live birth rates between the two arms was less than 2% once 6

or more blastocysts were available. For women aged 38 to 40, the difference in live birth rates

favoured IVF alone with 6 or fewer blastocysts and differed by 2% or less once 7 or more blastocysts

were available. For women aged 41 to 42, IVF was approximately 3% more effective than IVF/PGS

alone when 3 or fewer blastocysts were available and differed by 2% or less when 5 to 8 blastocysts

were available. With 9 or 10 blastocysts, IVF/PGS was 3.1% and 4.5% more effective than IVF

alone, respectively. For women greater than 42 years of age, the difference in live birth rates between

the two groups was less than 3% regardless of the number of blastocysts available.

32

4.1.3 Incremental Cost-Effectiveness Ratios

For women under age 35, the ICER fell into the northwest quadrant (meaning that IVF/PGS was

dominated by IVF alone, see Section 1.5 and Figure 8.1) in all cases regardless of the number of

blastocysts available. For women aged 35 to 37, the ICER fell into the northwest quadrant when

seven or fewer blastocysts were available and in the northeast quadrant (meaning that IVF/PGS was

more costly and more effective than IVF alone) when eight or more blastocysts were available. For

women aged 38 to 40, the ICER fell into the northwest quadrant when eight or fewer blastocysts

were available and in the northeast quadrant when nine or ten blastocysts were available. For women

aged 41 to 42, the ICER fell into the northwest quadrant when six or fewer blastocysts were available

and in the northeast quadrant when seven or more blastocysts were available. For women over age

42, the ICER fell into the northwest quadrant when one, two or three blastocysts were available, in

the southwest quadrant (meaning that IVF/PGS was less costly and less expensive than IVF alone)

when four to eight blastocysts were available and in the southeast quadrant (meaning that IVF/PGS

was less costly and more effective) when nine or ten blastocysts were available. The actual ICER

values are shown in Table 7.4.

4.1.4 Sensitivity Analyses

4.1.4.1 One Way Sensitivity Analysis

Figure 8.5 shows the results of one-way sensitivity analyses, in the form of tornado diagrams, for

each of the five age groups and for blastocyst cohort sizes of one, five and ten. For women less than

age 35, the model was most sensitive to the implantation rates of IVF alone and IVF/PGS. It was also

sensitive to the cost of embryo biopsy and PGS. It was less sensitive to the rates of EPL with IVF

alone and IVF/PGS and to the dropout rate. These findings were true regardless of the number of

blastocysts available.

For women aged 35 to 37, the model was most sensitive to the implantation rates of IVF alone and

IVF/PGS when fewer blastocysts were available. When the blastocyst size increased to ten, the

model was most sensitive to the dropout rate, following by the implantation rates and then the EPL

rates of IVF alone and IVF/PGS. For women aged 38 to 40, the model was again most sensitive to

the implantation rates of IVF alone and IVF with PGS regardless of the number of blastocysts

available. As with the younger age groups, the cost of embryo biopsy and PGS was a factor with

33

smaller blastocyst cohort sizes but not larger ones and, conversely, the dropout rate became a more

significant driver of the model results as the blastocyst cohort size increased.

For women aged 41 to 42, the model was most sensitive to the implantation rates of IVF alone and

IVF/PGS when the blastocyst cohort size was very small or very large. With five blastocysts, the

model was most sensitive to the EPL rate of IVF alone, followed next by the implantation rates of

IVF/PGS and IVF alone. For women older than age 42, the model was most sensitive to the

implantation rates of IVF/PGS and IVF alone when the cohort size was small, to the EPL rate of IVF

alone and the implantation rates of IVF/PGS and IVF alone when the cohort size was intermediate,

and to the dropout rate, followed by the EPL and implantation rates of both groups, when the cohort

size was largest.

The model was not sensitive to the costs of COH and OPU nor the costs of fertility medications. The

model was not sensitive to the costs of embryo cryopreservation and storage. The model was not

sensitive to productivity costs of either partner.

4.1.4.2 Probabilistic Sensitivity Analysis

The results of the PSA are shown quantitatively in Table 7.5 and graphically, with scatter plots by

age group, in Figure 8.6.

For women less than age 35, regardless of the number of blastocysts available, the majority of ICER

estimates fell into the northwest quadrant of the CE plane. More than 95% of ICER estimates fell

into the northwest quadrant when seven or fewer blastocysts were available. More than 80% of ICER

estimates fell into the northwest quadrant when eight or more blastocysts were available, with most

of the remaining estimates landing in the northeast quadrant.

For women aged 35 to 37, as the number of blastocysts increased, the ICER estimates began to

migrate from the northwest quadrant towards the northeast quadrant of the CE plane. With three or

fewer blastocysts, more than 90% of the ICER estimates fell into the northwest quadrant. With eight

or more, 36% or fewer of the ICER estimates fell into the northwest quadrant, while 64% or more

fell into the northeast quadrant.

For women aged 37 to 40, the ICER estimates migrated from the northwest to northeast quadrants as

the number of blastocysts increased. More than 95% of ICER estimates fell into the northwest

34

quadrant when three or fewer blastocyst were available. This number fell to 25% when ten

blastocysts were available, with the remainder of the estimates falling into the northeast quadrant.

For women aged 41 to 42 with three or fewer blastocysts, more than 87% of ICER estimates fell into

the northwest quadrant, with most of the remaining estimates falling into the northeast quadrant and

less than 1% falling into the southwest quadrant. In the case of four blastocysts, 77.7% of estimates

fell into the northwest quadrant, 16.6% into the northeast quadrant, 5.6% into the southwest quadrant

and 0.1% into the southeast quadrant. With five or more blastocysts, the ICER estimates spanned all

four quadrants to varying degrees, with an increasing percentage of ICER estimates falling into the

northeast and southeast quadrants as the number of blastocysts increased. With ten blastocysts, there

was a 46.4% chance that IVF/PGS was more effective and more costly than IVF alone and a 40.5%

chance that IVF/PGS was more effective and less costly than IVF alone.

For women older than age 42, more than 95% of ICER estimates fell into the northwest quadrant

when one or two blastocysts were available. This number decreased to 81.1% with three blastocysts.

Once four blastocysts were available, the majority (61.4%) of ICER estimates fell into the southwest

quadrant, although estimates did occupy all four quadrants. This was also true for the case of five

blastocysts. With six blastocysts available, the ICER estimates fell into the southwest quadrant 75%

of the time and into the southeast quadrant 25% of the time. With seven and eight blastocysts, these

numbers shifted to 66% and 33%, and 55% and 45% respectively. With nine or ten blastocysts, the

majority of ICER estimates fell into the southeast quadrant – 58% and 64% respectively.

4.2 Alternative Perspectives

Table 7.3 summarizes the specific costs included in each of the three alternative perspectives. The

point estimates for C, E and ICER by age, number of blastocysts and perspective can be found in

Table 7.6(a-c). Scatter plots of selected PSA results can be found in Figure 8.7(a-b).

4.2.1 Direct Health Care

When productivity costs were excluded from the model, the results were, overall, similar to those of

the societal perspective. IVF/PGS was dominated by IVF alone for women less than age 35

regardless of the number of blastocysts available. IVF/PGS was also dominated by IVF alone for

women with four or fewer blastocysts, regardless of age. IVF/PGS was more effective and more

costly than IVF alone for women aged 35 to 37 with eight or more blastocysts, aged 38 to 40 with

35

nine or more blastocysts, and aged 41 to 42 with seven or more blastocysts. ICER point estimates for

these scenarios ranged from $22,809 to $830,922 per live birth. For women age 42 and older,

IVF/PGS was less effective and less costly than IVF alone for those with five to eight blastocysts.

IVF/PGS dominated IVF alone in this age group when nine or ten blastocysts were available.

4.2.2 Hypothetical MOHLTC

This perspective assumed that the MOHLTC would cover the cost of PGS in addition to the health

care costs related to pregnancy or its complications. From this perspective, IVF/PGS was dominated

by IVF alone when six or fewer blastocysts were available regardless of age and for women under

age 35 regardless of blastocyst cohort size. IVF/PGS was more effective and more expensive than

IVF alone for women ages 35 to 37 with eight or more blastocysts, ages 38 to 40 with nine or more

blastocysts, ages 41 to 42 with seven or more blastocysts, and above age 42 with nine or more

blastocysts. The ICER point estimates for these scenarios with ICERs in the northeast quadrant

ranged from $91,696 to $1,032,609 per live birth. There were no scenarios in which IVF/PGS was

cost saving or in which IVF/PGS dominated IVF alone.

4.2.3 Patient Self-Pay

When the model was analyzed using only those health care costs not covered by the MOHLTC, the

results were, overall, similar to those of the direct health care perspective (Section 4.2.1). IVF/PGS

was dominated by IVF alone for women under age 35 and for women with four or fewer blastocysts

regardless of age. IVF/PGS was more effective and more costly than IVF alone for women aged 35

to 37 with eight or more blastocysts, aged 38 to 40 with nine or more blastocysts, and aged 41 to 42

with seven or more blastocysts. ICER point estimates for these scenarios ranged from $19,716 to

$828,586 per live birth. For women age 42 and older, IVF/PGS was less effective and less costly

than IVF alone for those with five to eight blastocysts. IVF/PGS dominated IVF alone in this final

age group when nine or ten blastocysts were available.

36

5 Discussion

5.1 Key Findings

This study found that IVF/PGS became increasingly more effective and less costly compared to IVF

alone as either the age of the woman or the number of available blastocysts increased. IVF alone,

however, was the dominant strategy for women under age 35 and for women with three or fewer

blastocysts regardless of age. Sensitivity analyses showed that the model was most sensitive to the IR

and EPL rates of IVF/PGS and IVF alone. When only direct health care costs or patient self-pay

costs were considered, the results were very similar to those obtained when considering all costs

from a societal perspective.

IVF/PGS was the dominant strategy over IVF alone (more effective and less costly) only when

advanced reproductive age was coupled with large numbers of blastocysts available for biopsy. This

is a very uncommon clinical scenario. Data from CARTR Plus / BORN Ontario showed that 79% of

cycles in Ontario in 2013 resulted in five or fewer usable blastocysts and 92% of cycles resulted in

eight or fewer usable blastocysts (111). Data from a large study by Franasiak et al., which excluded

those cases in which no blastocysts were available to biopsy, found that half of all of their PGS cases

had three or fewer “evaluable blastocysts” (8). They also found that the mean cohort size of

blastocysts suitable to biopsy was five or fewer for all women aged 34 and above (8).

The present study differed from most economic evaluations in that it did not compare two competing

treatment strategies but rather the impact of a supplemental intervention, namely PGS, on top of the

current treatment approach. PGS is therefore, by its very nature, associated with additional upfront

costs above and beyond those associated with IVF alone. It follows that for IVF/PGS to be less costly

than IVF alone overall, it must result in downstream consequences associated with fewer costs

compared to the downstream consequences of IVF alone. The two downstream consequences that are

best situated to fulfil this criteria are (1) a decrease in the number of FETs needed to achieve a

pregnancy or exhaust an embryo cohort, and (2) a decrease in the EPL rate. The current study found

that the incremental cost-effectiveness of IVF/PGS did in fact improve as the difference in EPL rate

became more pronounced between the two groups and as the number of embryo transfers necessary

to reach the end of a cycle increased.

This study found that the addition of PGS to IVF resulted in lower CLBRs compared to IVF alone in

the majority of scenarios. While this may seem counterintuitive, particularly given the widespread

37

adoption of PGS into clinical practice, it is not unexpected when the data used to inform our model

inputs for implantation and EPL rates are considered in greater detail. This data, obtained from

SART, does not support the magnitude of benefit, either in terms of increased IR or decreased EPL

rates, found in other published studies of PGS outcomes (79). For women less than age 35, there was

only a 3.5% absolute increase in IRs with IVF/PGS per embryo transfer. This increased to 11.5%,

18.4%, 29.4% and 32.9% for the subsequent age groups. Similarly, the absolute decrease in EPL

rates with a pregnancy achieved after the transfer of a euploid embryo compared to a pregnancy

achieved after the transfer of an unscreened embryo transfer was -0.7% in the youngest age group

and 2.6%, 6.9%, 15.2% and 17.3% in the subsequent age groups. These values contrast markedly to

those of, for example, Yang et al. who randomized 112 good prognosis patients under age 35 to

euploid SET versus unscreened SET at the blastocyst stage (51). They found that the IR increased by

25.1% (from 45.8% to 70.9%) and that the EPL rate decreased by 6.5% (from 9.1% to 2.6%) with

IVF/PGS compared to IVF alone (51). It is important to remember that these values represent the

outcomes of the first embryo transfer only and are not cumulative in nature.

The differences between SART and clinical trial data, including the study presented above, could be

explained in part by the fact that studies of PGS are often done on selected groups of fertility

patients, for example, those under 35 years old with a good prognosis or, conversely, those with

previously failed IVF cycles and/or RPL. It is also possible that the patients captured by SART who

underwent IVF/PGS cycles – particularly in the younger age groups – are not representative of the

infertility population as a whole either. Because PGS is not considered the standard of care for all

patients undergoing IVF, patients who underwent IVF/PGS, particularly young woman, may have

been fundamentally different in terms of prognostic factors compared with their equally young

counterparts who did not do PGS. They may have, for example, been more likely to have a history of

previously failed IVF cycles or recurrent implantation failure. This may partly explain the lower than

expected IRs with IVF/PGS in the younger age groups compared with the RCTs of PGS. On the

other hand, patients are often counseled to opt into PGS only if a minimum number of blastocysts are

available. Because women of more advanced reproductive ages are less likely to achieve this number

of blastocysts, it is possible that the patients in the older age groups of the SART database who

underwent IVF/PGS were of better prognosis than their counterparts who underwent IVF alone.

All three currently published RCTs of PGS, and the majority of observational studies, report only on

the results of the first embryo transfer rather than the CLBR from one fresh cycle start – a more

38

clinically meaningful outcome (123,124). There is strong evidence that IVF/PGS is superior to IVF

alone when comparing the IRs of a euploid to an unscreened embryo (62); however, it is much less

clear whether or not this advantage holds true when one compares the CLBR from all usable

embryos obtained from one fresh cycle start. Moreover, many of these studies included the transfer

of multiple embryos at a time in one or both groups. Data from those studies are particularly difficult

to apply to Ontario where the government has instituted a very strict single embryo transfer policy

tied to its recent funding of IVF (15) and where the rate of eSET is already much higher than it is in

the United States (125,126). In addition to these issues with study design, there is a growing body of

evidence that the absolute benefits of PGS may not be as impressive as the original studies suggested

(62,127–130).

Importantly, the per transfer IR and EPL rates are not the only key driver of the CLBR after PGS –

either in our model or in reality. The number of available euploid blastocysts and the percentage of

transferrable euploid embryos relative to the size of the unscreened embryo cohort (i.e. the

aneuploidy rate) are also a key driver of the primary outcome of interest and factors that are often

overlooked in the literature. Our study found that the model became sensitive to the dropout rate as

the blastocyst cohort size increased, particularly in the older age groups. This supports the hypothesis

that a smaller cohort of higher potential blastocysts can decrease dropout rates and therefore improve

the CLBRs. A higher IR per individual embryo transfer also has the ability to decrease the time to

conception, which although not explicitly studied in this model, is extremely important to patients.

5.2 Comparison with Other Studies

As highlighted in section 2.3.3, there are a limited number of other studies that have examined the

cost-effectiveness of PGS using CCS (68–73). These studies have used various methodologies,

looked at different populations of patients, and have had mixed results. In a published abstract, Neal

et al. reported on the costs of performing sequential single euploid versus unscreened embryo

transfers from one cycle of COH and OPU (71). They use individual patient data from a retrospective

cohort of nearly 12,000 IVF/PGS cycles. Unlike the present study, however, they calculated the

direct ART costs associated with each of those IVF/PGS cycles and compared them to the

hypothetical direct IVF costs and transfer outcomes should those cycles have proceeded without

PGS. Their end point was either live birth or exhaustion of the embryo cohort. Neither the average

female age nor the average blastocyst cohort size was described in the abstract. They found that there

were more than double the number of embryo transfers and many more pregnancy losses in the IVF

39

alone arm compared to the PGS arm. They concluded that, overall, the average cost until either live

birth or exhaustion of the embryo cohort was reached was lower with IVF/PGS and that IVF/PGS

was associated with a lower ACER when three or more embryos were available to biopsy (71). The

study did not model the option of dropping out of treatment prematurely, which would have further

favoured IVF/PGS over IVF alone. This study is limited by the fact that it relied upon a population of

patients who not only planned but ultimately completed a PGS cycle. It is reasonable to assume that

this population was different from the general IVF population and therefore the aneuploidy rates

(which were used to calculate the LBRs in their model) may not have been appropriate in their IVF

alone arm. Nonetheless, this study lends strong supports to the widely held belief that PGS can

reduce the number of transfers and therefore the time to cycle completion (either live birth or

exhaustion of embryo cohort) compared to IVF alone. It is difficult to assess, however, the

appropriateness of their included costs and clinical input parameters (particularly their LBR values)

from an abstract alone. Ultimately, the differing results between the Neal et al. abstract and our

present study are most likely due to a combination of differences in key clinical input parameters,

included costs and model structure.

The abstract published by Salem et al., on the other hand, used a model structure and reported results

that were more in keeping with the results of the present study (72). Their economic evaluation was

limited to women aged 39 and up and focused on the impact of varying PGS false positive rates on

the average cost per live birth. They found that even with low false positive rates, IVF/PGS resulted

in lower LBRs per cycle and therefore was less cost-effective than IVF alone (72). Their results, like

ours, were in large part due to the use of SART data to inform their clinical outcome parameters. One

major limitation of their study, however, was the assumption that all women had two blastocysts

suitable to biopsy rather than creating a model that incorporated varying blastocyst cohort sizes.

Scriven et al. recently published the only full length paper on the cost-effectiveness of PGS using

CCS (73). The study used Microsoft Excel to model sequential single embryo transfers from a single

cycle of COH, to a maximum of ten, until either a live birth was achieved or no embryos remained.

He used data from a study by Scott et al. (74) to arrive at LBRs of 25.4% and 43.1% for IVF alone

and IVF/PGS, respectively. Similarly, EPL rates of 8.5% and 5.1% for IVF alone and IVF/PGS,

respectively, were used. An embryo survival rate of 94% was used for cryopreserved embryos. The

study found that IVF/PGS was effective in decreasing the number of transfers and the EPL rate. It

also showed that IVF/PGS was effective at increasing the LBR from the first transfer. As expected

40

and in line with the results of Salem et al. (72), the Scriven study found that the greater the false

positive rate of PGS, the more the model favoured IVF alone. His model also found that regardless of

embryo cohort size, IVF/PGS resulted in fewer live births compared to IVF alone and was therefore

not cost-effective from a societal perspective. This study did not model dropout rates nor did it

include productivity costs associated with ART treatment itself. Furthermore, the study used data

from a single small non-selection study of PGS using aCGH. It does, however, strongly support the

argument that until PGS can be shown to be effective in increasing the CLBR from one fresh cycle of

IVF, it cannot be cost-effective.

5.3 Strengths

For this study, a novel decision model using TreeAge Pro to study the CLBRs from one fresh cycle

of IVF with or without the addition of PGS was designed. This model was specifically created to be

easily adaptable to varying blastocyst cohort sizes and to reflects the clinical reality that the number

of transferrable embryos after PGS is a function of both female age and original embryo cohort size.

Moving forwards, this model could be modified to study other interventions pertaining to IVF – for

example, embryo selection using time-lapse imaging or mitochondrial DNA content. It could also be

updated with new clinical and cost parameters as they become available.

This model is not only novel in its structure but it also includes a number of features absent from

other models used to study cost-effectiveness in the field of infertility. Not only were dropout rates

included in the model, but differential dropout rates were used depending on if an embryo transfer

failed to result in a pregnancy or if it resulted in a pregnancy that was subsequently lost. To our

knowledge, we are not aware of any other studies to date that have included this feature. The present

study did not assume that all EPLs were failed intrauterine pregnancies treated with surgical

management. Costs were divided between expectant, medical and surgical management options for

intrauterine EPLs and between medical and surgical management options for ectopic pregnancies.

Finally, the productivity costs of the partner were included in our study – a societal cost that has been

excluded from the majority of previously published CEAs related to fertility treatments.

The use of population-based data to inform our model means that the results of this study are more

generalizable to the larger infertility population than cost-effectiveness studies that have relied on the

results of a single trial for their model inputs. This, together with the extensiveness of the model

41

design and included costs, makes this study one of the most comprehensive studies of the cost-

effectiveness of PGS to date.

5.4 Limitations

5.4.1 Clinical Parameters

There were a number of limitations in this study related to the values used to inform implantation and

EPL rates with IVF/PGS and IVF alone. The pace at which PGS technology is advancing has made it

challenging for administrative databases to keep up with complete and accurate data collection. The

CARTR Plus / BORN Ontario database does not currently capture the ploidy status of transferred

embryos nor does it accurately or consistently differentiate screened from unscreened embryos at the

time of embryo transfer. For these reason, we were unable to use Ontario data to inform the

implantation and EPL rates in our model (see Section 6.5). Instead this study relied upon population

level data from SART; however, because individual level data was not available, it was not possible

to control for potential differences between the group of patients who underwent IVF/PGS and those

who underwent IVF alone.

Additionally, this study used constant implantation and pregnancy loss rates that were independent of

the number of previous transfers performed. This was done because data on IRs and EPL rates by

number of previous unsuccessful transfers were not available. There is reason to believe, however,

that the LBR may decrease with each successive transfer rather than remaining constant. Similarly,

the present study used a constant dropout rate after failed transfers and a constant, but slightly higher,

dropout rate after EPLs, regardless of the number of previous transfers or remaining transferrable

embryos. The same dropout rates were used in both arms of the model. To our knowledge, there is no

available data to either support or refute these specific assumptions. A dropout rate that increased

with the number of previous unsuccessful transfers would have favoured the IVF/PGS arm of the

model because subjects in that arm underwent fewer total embryo transfers before exhausting their

embryo cohort. To address these limitations, the dropout rates, along with the IR and EPL rates, were

subjected to one-way and probabilistic sensitivity analyses (see Section 4.1.4). Furthermore, the

model was designed such that it can be updated in the future and reanalyzed when these data become

available.

As discussed in section 2.1.3, CCS can be performed using a number of different technologies,

including aCGH and NGS. Much of the currently available clinical data – both in population

42

databases and in published trials – is derived from aCGH-based technologies. Recently, however,

there has been a rapid shift in the global landscape towards NGS-based approaches. Studies

comparing the two technologies (and using NGS as the gold standard) have repeatedly found that

aCGH is associated with a higher false positive rate than initially believed – meaning that aneuploid

(and particularly mosaic aneuploid) embryos were unknowingly being misclassified as euploid and

subsequently transferred (118,131). Mosaic embryos are less likely to implant and more likely to lead

to pregnancy losses compared with euploid embryos (132). This may be one of the reasons why PGS

outcome data from population-based sources such as SART reported differences in EPL rates

between screened and unscreened embryos that were less than expected based on earlier small-scale

studies.

5.4.2 Cost Data

In addition to the limitations described above, there were a number of limitations related to the cost

estimates used in this study. Productivity losses were estimated using the best available data;

however, detailed information relating to the amount of time absent from work due to a fresh IVF

cycle, a frozen embryo transfer cycle, a normal pregnancy, a pregnancy resulting in an intrauterine

EPLs or a pregnancy resulting in an ectopic pregnancy was not available. The data that were

available were not specific to Canada. Productivity losses were therefore estimated using expert

opinion and average hourly wage rates from Statistics Canada. Even less data were available relating

to the productivity losses of partners of individuals going through IVF procedures. We chose to

include these costs, nonetheless, to reflect the clinical reality that the IVF process often results in

work absences for two people rather than one. We stopped short, however, of extending the partners’

productivity losses into pregnancy and delivery. Sensitivity analyses, including the alternative

perspectives analyzed, found that changes in productivity loss estimates did not significantly impact

the results; therefore, it is unlikely that the inclusion of additional productivity losses on behalf of the

partner would have impacted our conclusions.

Cost estimates relating to the management of EPL were limited by the lack of information on the

relative use of expectant management, medical management and surgical management to treat EPL

in Canada. Other studies have chosen to assume that all cases of EPL were managed surgically. This

would overestimate the costs associated with EPL and so in the present study, we used a conservative

estimate and distributed costs equally amongst all three management options. Hospital costs were

obtained from OCCI and as a result, cost estimates for surgical management of EPL and medical and

43

surgical management of ectopic pregnancies did not account for the proportion of those conditions

managed successfully in an outpatient setting. IVF medication costs were calculated based on expert

opinion and local practice; however, it is acknowledged that there are many accepted protocols for

COH, FET cycles, and luteal phase support and that each require different medications for varying

lengths of time. To account for these and other uncertainties, extensive one-way and probabilistic

sensitivity analyses were performed. Ultimately, IR and EPL rates, and to a lesser degree dropout

rates, were found to be much more significant drivers of the model outcome than the costs related to

EPL management or IVF medications.

5.4.3 Model Design

The final group of limitations that must be addressed are those related to the model design itself.

While the present model is more representative and comprehensive than other models of the CLBR

after IVF/PGS, it is not perfect. The model does not include the option of ongoing aneuploid

pregnancies nor the related costs of prenatal diagnosis and genetic termination in either arm. In

reality, IVF/PGS detects embryos with viable aneuploidies and excludes them from the transferrable

cohort. IVF/PGS should therefore result in lower rates of ongoing aneuploid pregnancies and genetic

terminations compared to IVF alone. The costs associated with these pregnancies were not captured

in the present study. In this study, it was assumed that all usable blastocysts were successfully

cryopreserved and thawed without any attrition. Additionally, the model assumed that all transfers

were single embryo transfers and therefore did not capture the benefits – either with respect to health

outcomes or cost savings to the health care system – of a lowering of the multiple birth rate with

IVF/PGS compared to IVF alone.

5.5 CARTR Plus / BORN Ontario

At the outset of this study, the intent was to use data captured in CARTR Plus / BORN Ontario to

inform the clinical input parameters of our model. However, because of a number of key limitations

with respect to how information related to PGS were captured, data was ultimately obtained

elsewhere. It is worthwhile, therefore, to take a moment to discuss medical registries in general, and

CARTR Plus / BORN Ontario specifically, in order to explore ways in which CARTR Plus / BORN

Ontario may consider adapting in order to better meet the needs of its end users.

This analysis builds on the work done by Arts et al. in their 2002 review and synthesis of the

literature surrounding data quality in medical registries (133). In their publication, Arts et al. put

44

forth a categorization of causes of insufficient data quality and propose a framework of procedures in

order to assure high quality data in medical registries. They organized the causes of poor data quality

into three time points (set up of the registry, data collection and quality improvement) and two

locations (central coordinating centre and local sites). Specific errors are further categorized as either

random or systematic in nature. Their complete categorization is shown in Appendix 9.1.

A medical registry can be defined as a “systematic collection of a clearly defined set of health and

demographic data for patients with specific health characteristics, held in a central database for a

predefined purpose” (133). Like other registries, birth registries can be used for several purposes

including to improve patient care and conduct research. In order to be effective, however, its data

must be of a sufficient quality for the needs of its intended users (133). Data quality is generally

considered a function of two constructs – accuracy, the extent to which the data is true, and

completeness, the extent to which all data that could have been included is included (133).

In 2009, with funding from the Ontario MOHLTC, the Ontario Perinatal Surveillance System was

combined with five other partner groups to form BORN Ontario (134). Its mission includes being “an

authoritative source of accurate, trusted and timely information” about maternal, newborn and

childhood health in Ontario (135). In 2012, the medical directors of Canadian fertility clinics entered

into an agreement with BORN Ontario to have it “collect, analyze and report” the data contained

within the CARTR, which was subsequently rebranded as CARTR Plus. This joint undertaking

allows for data from Ontario ART patients to be linked with their pregnancy, birth and childhood

data collected through BORN Ontario (78).

5.5.1 Data Quality

The most significant challenge of working with the individual-level data obtained from CARTR Plus

/ BORN Ontario concerned the identification of individuals undergoing IVF/PGS cycles. In order to

study the effects of PGS (or any other intervention) it is of critical importance to accurately identify

which subjects within the cohort were exposed to the intervention (the treatment group) and which

were not (the control group). Inclusion of too many controls in the treatment group can lead to an

underestimation of treatment effect, or vice versa. When CARTR Plus was set up, the importance of

capturing which cycles included PGS and/or PGD was recognized and these terms were clearly

defined in their data dictionary (available at http://datadictionary.bornontario.ca/encounters-and-

alphabetical-lists/fertility-cartr-plus/). During the design phase of the registry, however, a decision

45

was made to capture this information under the data point “Reason for Treatment Cycle” – a pick list

of twenty items relating to the underlying causes of infertility or reasons for ART. Those options are:

- Male factor

- Endometriosis

- Tubal factor

- PCOS

- Other ovulatory disorder

- Diminished ovarian reserve

- Advanced female age

- Uterine factor

- Peritoneal factor or severe adhesions

- Gonadotoxic therapy

- Other female factor

- No male partner

- No female partner

- Unexplained infertility

- PGD for known genetic factor

- PGS for aneuploidy screening

- Oocyte banking - cancer treatment

- Oocyte banking - social reasons

- Embryo banking - cancer treatment

- Embryo banking - social reasons.

Users can choose as many items from this list as are applicable for a given case.

The primary problem with capturing PGS cycles in this way is that PGS in and of itself is not a

reason for someone to pursue IVF. PGS is better thought of as an additional test that is available to

those people undergoing IVF for another reason. It follows that PGS should be recorded in a method

similar to other tests, tools or procedures such as ICSI, assisted hatching or time-lapse photography.

One possible solution is to separate this pick-list into two separate data points, one entitled “fertility

diagnoses” and one “other indications for treatment” and to remove PGS from this section altogether

and record it elsewhere. Under this proposed scheme, the first list (“Reasons for Treatment cycle -

Fertility Diagnoses”) could include the following:

- Male factor

- Endometriosis

- Tubal factor

- PCOS

- Other ovulatory disorder

- Diminished ovarian reserve

- Advanced female age

- Uterine factor

- Peritoneal factor or severe adhesions

- Gonadotoxic therapy

- Unexplained infertility

46

- Other female factor

- No known fertility issue.

The second list (“Reasons for Treatment Cycle - Other Reasons for Treatment”) could include:

- No male partner

- No female partner

- PGD for known genetic factor

- Oocyte banking – oncology

- Oocyte banking – non-oncology medical

- Oocyte banking – non-medical

- Embryo banking – oncology

- Embryo banking – non-oncology medical

- Embryo banking – non-medical.

Users could choose as many options from each of the two lists as are applicable for a given case.

PGS could be selected regardless of which options are selected as the reasons for treatment. These

changes would allow for data about the reasons for treatment to be collected in a more informative

way.

The second significant issue that limited the ability to accurately differentiate treatment subjects from

controls related to the way in which the data was entered. Although the definition of PGS was clearly

defined in the dictionary, it was unclear to those entering the data how to handle cases in which PGS

was planned but not ultimately performed (T. Nichols, personal communication). It is known that

these cases make up a significant portion of all planned PGS cases because patients and physicians

often set a target number of blastocysts a priori above which PGS will be performed and below

which it will not – and that target is not always met. Without clear instructions about how to input

these cases, there is reason to believe that some users selected “PGS” from the “Reason for

Treatment Cycle” pick-list for all cases in which PGS was planned whereas others selected it only for

those cases that ultimately reached the point of embryo biopsy. There is also the issue of capturing

those cycles in which a fresh embryo transfer was performed and only on surplus blastocysts undergo

PGS. To overcome this issue, CARTR Plus is encouraged to add two data points to capture: (1)

whether or not PGS was planned at the start of an IVF cycle (on all blastocysts or excess blastocysts

only) and (2) whether or not embryo biopsy for the purposes of PGS was ultimately performed. If

PGS was planned but not performed, there could be a pick-list of reasons why it was not – similar to

the “Reason for no ET” data point. Options could include:

- Cycle cancelled prior to OPU

- No oocytes retrieved

- All embryos arrested

- No embryos suitable to biopsy

47

- Unscreened embryo(s) transferred and/or cryopreserved without biopsy

- Other.

A similar system could be put in place for PGD, with the exception that PGD should also remain as

an option under the data point “Reasons for Treatment Cycle – Other Reasons for Treatment.”

The above suggestions would allow data users to accurately differentiate three groups of patients:

those who did not do PGS, those who planned to but did not do PGS, and those who did PGS. And

while this would eliminate one of the major issues around using CARTR Plus / BORN Ontario for

PGS and/or PGD-related research, these changes alone would not be sufficient. That is because the

registry does not currently capture the genetic results of PGS and/or PGD cycles. Until this

information is captured accurately and completely, the utility of CARTR Plus / BORN Ontario for

PGS and/or PGD-related research will remain appreciably compromised. The number of PGS and/or

PGD cycles is rising rapidly and the complete and accurate recording of these results is becoming

increasingly important. In 2012, there were 165 fresh or frozen PGS and/or PGD cycles in Canada

(136). In 2014 that number rose to 762 (126) and in 2015 it jumped to 2,355 (5). CARTR Plus /

BORN Ontario is encouraged to act swiftly to update their registry to include embryo biopsy results

as soon as is feasible. The information captured should include, at a minimum, the following:

- day of embryo biopsy (i.e. Day 3, 5 or 6)

- whether or not the oocytes and/or embryos had been previously frozen prior to biopsy

- platform used (i.e. FISH, aCGH, qPCR, NGS)

- biopsy results (including cases of mosaicism, indeterminate and failure to amplify).

Promisingly, CARTR Plus / BORN Ontario has already updated the data point “Reason for no ET” to

reflect the clinical reality that many freeze-all cycles are a result of PGS. Furthermore, CARTR Plus /

BORN Ontario plans to start reporting CLBR data in 2017 (137). A summary of some of the causes

of inadequate data quality found in CARTR Plus / Born Ontario along with suggestions for

improvement can be found in Table 7.7.

5.5.2 Data Accessibility

For a registry to succeed in achieving its goals the data have to be of good quality (accurate and

complete) and also accessible to its end users. Those end users include not only the stakeholders

involved in a registry’s creation but may also government agencies, policy-makers, academic

institutions, community organizations and individual clinicians and researchers. And just as data

quality can be understood as being a function of accuracy and completeness, data accessibility can be

48

thought of as a function of both timeliness and ease of access. These two constructs are interrelated

because overly bureaucratic access procedures can hinder timeliness. Put together, all these factors

form the basis of data usability. A graphical representation of the components of data usability is

shown in Figure 8.8.

According to its mission, BORN Ontario strives to “mobilize information and expertise to optimize

care” (135). Working with researchers across the province and encouraging high quality research

using their data is consistent with this mission. Therefore, in order to increase the usability of their

data to individual researchers, BORN Ontario may consider improving some features of their data

accessibility. While recognizing that the privacy of Ontarians and the security of personal health

information is paramount, timely access to data is necessary to perform truly impactful research. This

is particularly true in the field of assisted reproductive technologies, where new technologies are

continuously emerging. PGS is a particularly good example of this phenomenon. Only five or six

years ago, PGS was being performed overwhelmingly using FISH technology. A study published

now using PGS data from five years ago, therefore, would be entirely irrelevant in today’s clinical

reality. The same argument could even be made, but to a lesser degree, for the transition that is

currently underway from aCGH to NGS-based CCS. CARTR Plus / BORN Ontario has the potential

to support countless research endeavors and influence numerous policy debates on topics spanning

from the preconception period all the way through early childhood. To realize its goals, however, it

should continue to work towards improving their procedures for efficient and timely access to their

data. They are also encouraged to put in place mechanisms to facilitate end-user feedback and use

that feedback to inform their quality improvement plans. Given the paucity of published population

level data on PGS outcomes, CARTR Plus / BORN Ontario has the ability to uniquely position

themselves to provide answers to critical questions for years to come. I would welcome the

opportunity to work with them towards this goal both now and in the future.

5.6 Non-Economic Advantages of PGS

Cost-effectiveness analyses and other economic evaluations are important tools for quantifying the

relative trade-offs between alternative treatment strategies. However, when it comes to clinical

decision-making, particularly on an individual-patient level, there are other factors to consider when

deciding whether or not to adopt a new technology like PGS. One of the main limitations of the

present study is its inability to measure either the importance of time to conception or the

psychological benefits of reduced duration of infertility. Furthermore, there are also clear

49

psychological benefits to minimizing the incidence of EPL, especially among those already

struggling to conceive. Whether or not fertility patients would be willing to accept a slightly lower

CLBR with the addition of PGS in order to minimize their risk of a pregnancy loss and/or increase

their chance of achieving a pregnancy sooner is unknown and deserves further study. Additionally,

the potential psychological benefits of reduced time to conception and lower EPL rates may also

translate to economic benefits if one assumes that people suffering from infertility are not able to

contribute maximally to society due to the negative effects of their disease. These potential benefits

are not currently being captured in CEAs of PGS. It is important to note, conversely, that a

significant proportion of patients who choose to do PGS will have no euploid embryos available to

transfer and that further study is also required regarding the psychological effects of this outcome.

5.7 Future Directions

This study has highlighted a number of gaps in the literature that could and should be addressed

moving forwards. It has also become clear that this model has great potential to be used in other

capacities to continue to study the cost-effectiveness of PGS. This decision model comparing

IVF/PGS to IVF alone can and should be updated as new clinical and cost data become available.

Specifically, as data on pregnancy and live birth rates after PGS using NGS are published, this model

could be updated and the results reanalyzed with those values. The model could also be expanded to

address a number of the limitations discussed above including the possibility of ongoing aneuploid

pregnancies and of embryo non-survival after cryopreservation.

The process of collecting data for this study has drawn attention to a number of other potential

research opportunities. There is currently a lack of Canadian data regarding productivity losses

around fertility treatment. A study that collected information about time absent from work, for both

women and their partners, as applicable, due to infertility treatments and resulting normal and

abnormal pregnancies would be extremely valuable, not only for economic evaluations related to

PGS but for numerous economic evaluations related to the treatment of infertility in general. A study

that quantified the dropout rates within a single fresh IVF cycle, both with and without the use of

PGS, would also be an extremely valuable addition to the scientific literature and could easily be

done using retrospective records.

Finally, as discussed above, there are many opportunities to study the benefits of PGS that are not

directly addressed by way of economic evaluation including time to conception, time to cycle

50

completion and reduction in EPL rates. Methodologies should be continually developed to

incorporate these values and preferences into future studies of PGS and other embryo selection

techniques.

5.8 Conclusions

In this study, we constructed a novel decision tree to examine the cost-effectiveness of IVF/PGS

compared to IVF alone, in terms of live births, from a societal perspective in Ontario, Canada. The

study found that IVF/PGS became increasingly more effective and less costly compared to IVF alone

as either the age of the woman or the number of available blastocysts increased. IVF alone, however,

remained the dominant strategy for women under age 35 and for women with three or fewer

blastocysts regardless of age. IVF/PGS was the dominant strategy only when advanced reproductive

age was coupled with large numbers of blastocysts available for biopsy – a very uncommon clinical

scenario. Therefore, currently available data does not support the use of IVF/PGS as a cost-effective

intervention except in older women with large embryo cohorts. Because the model was found to be

sensitive to the IR and EPL rates of both IVF/PGS and IVF alone, high quality research is needed in

order to better understand the impact of PGS on CLBRs after IVF. The economic advantages of PGS

resulting from a reduction in the number of FETs and lower EPL rates should be studied alongside

other potential benefits of PGS, such as a reduction in the time to cycle completion and lower

dropout rates.

This study adds considerably to the small but growing body of literature concerning the cost-

effectiveness of PGS by studying the incremental cost-effectiveness of PGS as both a function of

female age and embryo cohort size. This novel, comprehensive model of PGS that we have built can

be used, and improved upon, to enhance our understanding of the potential benefits and drawbacks of

PGS. The results of this study can be used to encourage further discussions regarding the role of PGS

in clinical practice and to shape future research studies on the subject.

51

6 References

1. Bushnik T, Cook JL, Yuzpe AA, Tough S, Collins J. Estimating the prevalence of infertility in

Canada. Hum Reprod. 2012 Mar 1;27(3):738–46.

2. Cousineau TM, Domar AD. Psychological impact of infertility. Best Pract Res Clin Obstet

Gynaecol. 2007 Apr;21(2):293–308.

3. Cohen J. From Pythagoras and Aristotle to Boveri and Edwards: a history of clinical embryology

and therapeutic IVF. In: Textbook of Clinical Embryology [Internet]. Cambridge University Press;

2013. Available from: http://dx.doi.org/10.1017/CBO9781139192736.021

4. Fritz MA, Speroff L. Clinical gynecologic endocrinology and infertility [Internet]. 8th ed.

Philadelphia: Wolters Kluwer Health/Lippincott Williams & Wilkins; 2011.

5. Better Outcomes Registry & Network Ontario. Canadian Assisted Reproductive Technologies

Register (CARTR) Plus: Final treatment cycle and pregnancy outcome data for 2014 &

preliminary treatment cycle data for 2015. CFAS 62nd Annual Meeting; 2016 Sep; Toronto,

Ontario.

6. Wilton L. Preimplantation genetic diagnosis for aneuploidy screening in early human embryos: a

review. Prenat Diagn. 2002 Jun 1;22(6):512–8.

7. Nagaoka SI, Hassold TJ, Hunt PA. Human aneuploidy: mechanisms and new insights into an age-

old problem. Nat Rev Genet. 2012 Jul 1;13(7):493–504.

8. Franasiak J, Forman E, Hong K, Werner M, Upham K, Treff N, et al. The nature of aneuploidy

with increasing age of the female partner: a review of 15,169 consecutive trophectoderm biopsies

evaluated with comprehensive chromosomal screening. Fertil Steril. 2014 Mar;101(3):656–63.

9. Capalbo A, Rienzi L, Cimadomo D, Maggiulli R, Elliott T, Wright G, et al. Correlation between

standard blastocyst morphology, euploidy and implantation: an observational study in two centers

involving 956 screened blastocysts. Hum Reprod. 2014 Jun 1;29(6):1173–81.

10. Dahdouh EM, Balayla J, García-Velasco JA. Comprehensive chromosome screening improves

embryo selection: a meta-analysis. Fertil Steril. 2015 Dec;104(6):1503–12.

11. Forman EJ. Obstetrical and neonatal outcomes from the BEST Trial: single embryo transfer with

aneuploidy screening improves outcomes after in vitro fertilization without compromising delivery

rates. 2014;210(2):157.e1-157.e6.

12. Practice Committee of American Society for Reproductive Medicine. Multiple gestation associated

with infertility therapy: an American Society for Reproductive Medicine Practice Committee

opinion. Fertil Steril. 2012 Apr;97(4):825–34.

13. Practice Committee of the Society for Assisted Reproductive Technology and Practice Committee

of the American Society for Reproductive Medicine. Elective single-embryo transfer. Fertil Steril.

2012 Apr;97(4):835–42.

14. Quebec Commissaire a la sante et au bien-etre. Summary Advisory on Assisted Reproduction in

Quebec [Internet]. 2014 Jun [cited 2014 Oct 9]. Available from:

52

http://www.csbe.gouv.qc.ca/fileadmin/www/2014/Procreation_assistee/CSBE_PA_SummaryAdvi

sory_2014.pdf

15. Greenblatt E, Advisory Process for Infertility Services. Advisory Process for Infertility Services

Key Recommendations Report [Internet]. 2015 Jun [cited 2016 Jan 20] p. 20. Available from:

http://health.gov.on.ca/en/public/programs/ivf/docs/ivf_report.pdf

16. Ubaldi FM, Capalbo A, Colamaria S, Ferrero S, Maggiulli R, Vajta G, et al. Reduction of multiple

pregnancies in the advanced maternal age population after implementation of an elective single

embryo transfer policy coupled with enhanced embryo selection: pre- and post-intervention study.

Hum Reprod. 2015 Sep 1;30(9):2097–106.

17. Practice Committee of the American Society for Reproductive Medicine. Guidance on the limits to

the number of embryos to transfer: a committee opinion. Fertil Steril. 2017 Apr;107(4):901–3.

18. Andrews JR, Lawn SD, Dowdy DW, Walensky RP. Challenges in Evaluating the Cost-

effectiveness of New Diagnostic Tests for HIV-Associated Tuberculosis. Clin Infect Dis Off Publ

Infect Dis Soc Am. 2013 Oct 1;57(7):1021–6.

19. Drummond MF, Sculpher MJ, Claxton K, Stoddart GL, Torrance GW. Methods for the Economic

Evaluation of Health Care Programmes. Fourth Edition edition. Oxford, United Kingdom ; New

York, NY, USA: Oxford University Press; 2015. 464 p.

20. Okun N, Teitelbaum M, Huang T, Dewa CS, Hoch JS. The price of performance: a cost and

performance analysis of the implementation of cell-free fetal DNA testing for Down syndrome in

Ontario, Canada. Prenat Diagn. 2014 Apr;34(4):350–6.

21. Detsky AS, Laupacis A. Relevance of Cost-effectiveness Analysis to Clinicians and Policy

Makers. JAMA. 2007 Jul 11;298(2):221–4.

22. Goodman C. Introduction to Health Technology Assessment. Bethesda, MA: National Library of

Medicine (US); 2014.

23. Hoch JS, Dewa CS. A clinician’s guide to correct cost-effectiveness analysis: Think incremental

not average. Can J Psychiatry. 2008;53(4):267.

24. Detsky AS, Naglie IG. A Clinician’s Guide to Cost-Effectiveness Analysis. Ann Intern Med. 1990

Jul 15;113(2):147–54.

25. Hoch JS, Briggs AH, Willan AR. Something old, something new, something borrowed, something

blue: a framework for the marriage of health econometrics and cost‐effectiveness analysis. Health

Econ. 2002;11(5):415–30.

26. Gray AM, Clarke PM, Wolstenholme J, Wordsworth S. Applied Methods of Cost-effectiveness

Analysis in Healthcare. 1 edition. Oxford University Press; 2010. 328 p.

27. Owens DK. Interpretation of Cost-Effectiveness Analyses. J Gen Intern Med. 1998

Oct;13(10):716–7.

28. Domar AD, Zuttermeister PC, Friedman R. The psychological impact of infertility: a comparison

with patients with other medical conditions. J Psychosom Obstet Gynaecol. 1993;14:45–45.

53

29. Handyside AH, Penketh RJA, Winston RML, Pattinson JK, Delhanty JDA, Tuddenham EGD.

Biopsy of human preimplantation embryos and sexing by DNA amplification. The Lancet. 1989

Feb 18;333(8634):347–9.

30. Handyside AH, Kontogianni EH, Hardy K, Winston RML. Pregnancies from Biopsied Human

Preimplantation Embryos Sexed by Y-Specific DNA Amplification. Nature. 1990 Apr

19;344(6268):768–70.

31. Handyside AH. Preimplantation genetic diagnosis after 20 years. Reprod Biomed Online. 2010

Sep;21(3):280–2.

32. Verlinsky Y, Cieslak J, Freidine M, Lvakhnenko V, Wolf G, Kovalinskaya L, et al. Pregnancies

following pre-conception diagnosis of common aneuploidies by fluorescent in-situ hybridization.

Mol Hum Reprod. 1995 Jul 1;1(5):265–9.

33. Practice Committee of the Society for Assisted Reproductive Technology, Practice Committee of

the American Society for Reproductive Medicine. Preimplantation genetic testing: a Practice

Committee opinion. Fertil Steril. 2008 Nov;90(5, Supplement):S136–43.

34. De Vos A, Van Steirteghem A. Aspects of biopsy procedures prior to preimplantation genetic

diagnosis. Prenat Diagn. 2001 Sep 1;21(9):767–80.

35. Harper JC, Sengupta SB. Preimplantation genetic diagnosis: State of the ART 2011. Hum Genet.

2012 Feb;131(2):175–86.

36. Munne S, Dailey T, Sultan KM, Grifo J, Cohen J. Diagnosing and preventing inherited disease:

The use of first polar bodies for preimplantation diagnosis of aneuploidy. Hum Reprod. 1995 Apr

1;10(4):1014–20.

37. Scott KL, Hong KH, Scott Jr. RT. Selecting the optimal time to perform biopsy for

preimplantation genetic testing. Fertil Steril. 2013 Sep;100(3):608–14.

38. Twisk M, Mastenbroek S, van Wely M, Heineman MJ, Van der Veen F, Repping S.

Preimplantation genetic screening for abnormal number of chromosomes (aneuploidies) in in vitro

fertilisation or intracytoplasmic sperm injection. Cochrane Database Syst Rev [Internet]. 2006.

39. Mastenbroek S, Twisk M, van der Veen F, Repping S. Preimplantation genetic screening: a

systematic review and meta-analysis of RCTs. Hum Reprod Update. 2011 Jul 1;17(4):454–66.

40. Scott RT, Upham KM, Forman EJ, Zhao T, Treff NR. Cleavage-stage biopsy significantly impairs

human embryonic implantation potential while blastocyst biopsy does not: a randomized and

paired clinical trial. Fertil Steril. 2013 Sep;100(3):624–30.

41. Harper JC, Harton G. The use of arrays in preimplantation genetic diagnosis and screening. Fertil

Steril. 2010 Sep 1;94(4):1173–7.

42. Frumkin T, Malcov M, Yaron Y, Ben-Yosef D. Elucidating the origin of chromosomal aberrations

in IVF embryos by preimplantation genetic analysis. Mol Cell Endocrinol. 2008 Jan 30;282(1–

2):112–9.

54

43. Barbash-Hazan S, Frumkin T, Malcov M, Yaron Y, Cohen T, Azem F, et al. Preimplantation

aneuploid embryos undergo self-correction in correlation with their developmental potential. Fertil

Steril. 2009 Sep;92(3):890–6.

44. Dokras A, Sargent IL, Ross C, Gardner RL, Barlow DH. Trophectoderm biopsy in human

blastocysts. Hum Reprod. 1990 Oct 1;5(7):821–5.

45. Kokkali G, Vrettou C, Traeger-Synodinos J, Jones GM, Cram DS, Stavrou D, et al. Birth of a

healthy infant following trophectoderm biopsy from blastocysts for PGD of β-thalassaemia major:

Case report. Hum Reprod. 2005 Jul 1;20(7):1855–9.

46. Treff N, Scott RT Jr. Methods for comprehensive chromosome screening of oocytes and embryos:

capabilities, limitations, and evidence of validity. J Assist Reprod. 2012 May;29(5):381–90.

47. Dahdouh E, Balayla J, Audibert F, Genetics Committee, Society of Obstetricians and

Gynaecologists of Canada. Technical update #323: preimplantation genetic diagnosis and

screening. J Obstet Gynaecol Can. 2015 May;37(5):451–63.

48. Brezina PR, Anchan R, Kearns WG. Preimplantation genetic testing for aneuploidy: what

technology should you use and what are the differences? J Assist Reprod Genet. 2016

Jul;33(7):823–32.

49. Hellani A, Abu-Amero K, Azouri J, El-Akoum S. Successful pregnancies after application of

array-comparative genomic hybridization in PGS-aneuploidy screening. Reprod Biomed Online.

2008;17(6):841–7.

50. Fragouli E, Alfarawati S, Daphnis DD, Goodall N -nek., Mania A, Griffiths T, et al. Cytogenetic

analysis of human blastocysts with the use of FISH, CGH and aCGH: scientific data and technical

evaluation. Hum Reprod. 2011 Feb 1;26(2):480–90.

51. Yang Z, Liu J, Collins GS, Salem SA, Liu X, Lyle SS, et al. Selection of single blastocysts for

fresh transfer via standard morphology assessment alone and with array CGH for good prognosis

IVF patients: results from a randomized pilot study. Mol Cytogenet. 2012;5:24.

52. Munné S. Preimplantation Genetic Diagnosis for Aneuploidy and Translocations Using Array

Comparative Genomic Hybridization. Curr Genomics. 2012 Sep;13(6):463–70.

53. Handyside AH. PGD and aneuploidy screening for 24 chromosomes by genome-wide SNP

analysis: seeing the wood and the trees. Reprod Biomed Online. 2011 Dec;23(6):686–91.

54. Treff NR, Tao X, Ferry KM, Su J, Taylor D, Scott Jr. RT. Development and validation of an

accurate quantitative real-time polymerase chain reaction–based assay for human blastocyst

comprehensive chromosomal aneuploidy screening. Fertil Steril. 2012 Apr;97(4):819–824.e2.

55. Treff N, Scott RJ. Four-hour quantitative real-time polymerase chain reaction-based

comprehensive chromosome screening and accumulating evidence of accuracy, safety, predictive

value, and clinical efficacy. Fertil Steril. 2013 Mar 15;99(4):1049–53.

56. Yang Y-S, Chang S-P, Chen H-F, Ma G-C, Lin W-H, Lin C-F, et al. Preimplantation genetic

screening of blastocysts by multiplex qPCR followed by fresh embryo transfer: validation and

verification. Mol Cytogenet. 2015 Jul 8;8(1):49.

55

57. Handyside AH. 24-chromosome copy number analysis: a comparison of available technologies.

Fertil Steril. 2013 Sep;100(3):595–602.

58. Handyside AH, Wells D. Single Nucleotide Polymorphisms and Next Generation Sequencing. In:

Gardner DK, Sakkas D, Seli E, Wells D, editors. Human Gametes and Preimplantation Embryos

[Internet]. Springer New York; 2013 [cited 2016 May 20]. p. 135–45.

59. Wells D. Next-generation sequencing: the dawn of a new era for preimplantation genetic

diagnostics. Fertil Steril. 2014 May 1;101(5):1250–1.

60. Dahdouh EM, Balayla J, Garcia-Velasco JA. Impact of blastocyst biopsy and comprehensive

chromosome screening technology on preimplantation genetic screening: a systematic review of

randomized controlled trials. Reprod Biomed Online. 2015 Mar;30(3):281–9.

61. Lee E, Illingworth P, Wilton L, Chambers GM. The clinical effectiveness of preimplantation

genetic diagnosis for aneuploidy in all 24 chromosomes (PGD-A): systematic review. Hum

Reprod. 2015 Feb 1;30(2):473–83.

62. Chen M, Wei S, Hu J, Quan S. Can Comprehensive Chromosome Screening Technology Improve

IVF/ICSI Outcomes? A Meta-Analysis. PLOS ONE. 2015 Oct 15;10(10):1–21.

63. Maheshwari A, McLernon D, Bhattacharya S. Cumulative live birth rate: time for a consensus?

Hum Reprod. 2015 Dec 1;30(12):2703–7.

64. Mersereau J, Plunkett B, Cedars M. Preimplantation genetic screening in older women: a cost-

effectiveness analysis. Fertil Steril. 2008 Sep;90(3):592–8.

65. Resetkova N, Tobler KJ, Kearns WG, Werner EF. In vitro fertilization (IVF) with 23-chromosome

pair preimplantation genetic screening (PGS) is cost effective to achieve a live birth compared to

IVF alone for recurrent pregnancy loss (RPL). Fertil Steril. 2013 Sep 1;100(3):S99–100.

66. Murugappan G, Ohno MS, Lathi RB. Cost-effectiveness analysis of preimplantation genetic

screening and in vitro fertilization versus expectant management in patients with unexplained

recurrent pregnancy loss. Fertil Steril. 2015 May;103(5):1215–20.

67. Murugappan G, Shahine LK, Perfetto CO, Hickok LR, Lathi RB. Intent to treat analysis of in vitro

fertilization and preimplantation genetic screening versus expectant management in patients with

recurrent pregnancy loss. Hum Reprod. 2016 Aug;31(8):1668–74.

68. Resetkova N, Tobler KJ, Werner EF, Kearns WG. In vitro fertilization (IVF) with 23-chromosome

pair preimplantation genetic screening (PGS) from trophectoderm biopsy is more cost effective to

achieve a live birth compared to IVF alone. Fertil Steril. 2014 Sep 1;102(3):e176.

69. Patounakis G, Sundheimer LW, DeCherney AH, Hill MJ. Preimplantation genetic screening (PGS)

of embryos prior to transfer: a cost analysis of single embryo transfers (SET) and double embryo

transfers (DET). Fertil Steril. 2014 Sep 1;102(3):e5.

70. Hodes-Wertz B, McCulloh DH, Grifo J. Preimplantation genetic screening is cost effective in cost

per delivery compared to routine in vitro fertilization. Fertil Steril. 2015 Sep 1;104(3):e278.

56

71. Neal S, Morin SJ, Franasiak JM, Juneau CR, Zhan Y, Scott RT. Single embryo transfer (SET)

following comprehensive chromosome screening (CCS) is more cost effective than unscreened

sequential SET. Fertil Steril. 2016 Sep 1;106(3):e19.

72. Salem W, Ho JR, Bendikson KA, Chung K, Paulson R. IVF patients over age 39 experience

decreased cost effectiveness and live birth rates with preimplantation genetic screening: a decision

analytic model and cost effectiveness analysis. Fertil Steril. 2016 Sep 1;106(3):e60–1.

73. Scriven PN. Towards a better understanding of preimplantation genetic screening and cumulative

reproductive outcome: transfer strategy, diagnostic accuracy and cost-effectiveness. AIMS Genet.

2016;3(3):177–95.

74. Scott RJ, Ferry K, Su J, Tao X, Scott K, Treff N. Comprehensive chromosome screening is highly

predictive of the reproductive potential of human embryos: a prospective, blinded, nonselection

study. Fertil Steril. 2012 Apr;97(4):870–5.

75. Scott RJ, Franasiak JM, Forman EJ. Comprehensive chromosome screening with synchronous

blastocyst transfer: time for a paradigm shift☆. Fertil Steril. 2014 Sep 1;102(3):660–1.

76. Dahdouh EM, Balayla J, García-Velasco JA. Preimplantation genetic screening using

comprehensive chromosome screening: evidence and remaining challenges. Hum Reprod. 2015

Jun 1;30(6):1515–6.

77. Husereau D, Drummond M, Petrou S, Carswell C, Moher D, Greenberg D, et al. Consolidated

Health Economic Evaluation Reporting Standards (CHEERS)—Explanation and Elaboration: A

Report of the ISPOR Health Economic Evaluation Publication Guidelines Good Reporting

Practices Task Force. Value Health. 2013 Mar;16(2):231–50.

78. CARTR Plus [Internet]. [cited 2016 Aug 16]. Available from:

https://www.bornontario.ca/en/partnership-projects/cartr-plus/

79. Society for Assisted Reproductive Technology. SART National Summary Report (2014)

[Internet]. 2016 [cited 2016 Aug 5]. Available from:

https://www.sartcorsonline.com/rptCSR_PublicMultYear.aspx?ClinicPKID=0#patient-first-

attempt

80. Brandes M, van der Steen JOM, Bokdam SB, Hamilton CJCM, de Bruin JP, Nelen WLDM, et al.

When and why do subfertile couples discontinue their fertility care? A longitudinal cohort study in

a secondary care subfertility population. Hum Reprod. 2009 Dec 1;24(12):3127–35.

81. Olivius C, Friden B, Borg G, Bergh C. Why do couples discontinue in vitro fertilization treatment?

a cohort study. Fertil Steril. 2004 Feb;81(2):258–61.

82. Gameiro S, Boivin J, Peronace L, Verhaak CM. Why do patients discontinue fertility treatment? A

systematic review of reasons and predictors of discontinuation in fertility treatment. Hum Reprod

Update. 2012 Nov 1;18(6):652–69.

83. Domar AD, Smith K, Conboy L, Iannone M, Alper M. A prospective investigation into the reasons

why insured United States patients drop out of in vitro fertilization treatment. Fertil Steril. 2010

Sep;94(4):1457–9.

57

84. De Vries MJ, De Sutter P, Dhont M. Prognostic factors in patients continuing in vitro fertilization

or intracytoplasmic sperm injection treatment and dropouts. Fertil Steril. 1999 Oct;72(4):674–8.

85. Schröder AK, Katalinic A, Diedrich K, Ludwig M. Cumulative pregnancy rates and drop-out rates

in a German IVF programme: 4102 cycles in 2130 patients. Reprod Biomed Online. 2004 Jan

1;8(5):600–6.

86. Health Technology & Policy Unit, School of Public Health, Department of Public Health Services.

Assisted Reproductive Technologies (ARTs) Final Report [Internet]. Edmonton, AB: University of

Alberta; 2014 Feb [cited 2014 Nov 20] p. 507. Available from:

http://www.health.alberta.ca/documents/AHTDP-Assisted-Reproductive-Technologies-2014.pdf

87. McDowell S, Murray A. Barriers to continuing in vitro fertilisation--why do patients exit fertility

treatment? J Obstet. 2011 Feb;51(1):84–90.

88. Huppelschoten AG, Dongen AJCM van, Philipse ICP, Hamilton CJCM, Verhaak CM, Nelen

WLDM, et al. Predicting dropout in fertility care: a longitudinal study on patient-centredness.

Hum Reprod. 2013 May 21;det236.

89. Macaldowie A, Lee E, Chambers G. Assisted reproductive technology in Australia and New

Zealand 2013 [Internet]. Sydney, Australia: National Perinatal Epidemiology and Statistics Unit,

University of New South Wales; 2015 Sep [cited 2016 Apr 29] p. 89. Available from:

https://npesu.unsw.edu.au/sites/default/files/npesu/data_collection/Assisted%20reproductive%20te

chnology%20in%20Australia%20and%20New%20Zealand%202013.pdf

90. Karvir H, Hunter Cohn K, Arredondo F, Miller B, Gutmann J, Leondires M, et al. Pregnancy loss

is a major predictor of premature IVF discontinuation. Pacific Coast Reproductive Society; 2016

Mar 9; Palm Springs, CA.

91. Al-Inany HG, Abou-Setta AM, Aboulghar MA, Mansour RT, Serour GI. HMG versus rFSH for

ovulation induction in developing countries: a cost–effectiveness analysis based on the results of a

recent meta-analysis. Reprod Biomed Online. 2006;12(2):163–9.

92. Fiddelers AAA, Dirksen CD, Dumoulin JCM, Montfoort APA van, Land JA, Janssen JM, et al.

Cost-effectiveness of seven IVF strategies: results of a Markov decision-analytic model. Hum

Reprod. 2009 Jul 1;24(7):1648–55.

93. Griffiths A, Dyer SM, Lord SJ, Pardy C, Fraser IS, Eckermann S. A cost-effectiveness analysis of

in-vitro fertilization by maternal age and number of treatment attempts. Hum Reprod. 2010 Apr

1;25(4):924–31.

94. Provoost V, Pennings G, Sutter PD, Gerris J, Velde AV de, Dhont M. To continue or discontinue

storage of cryopreserved embryos? Patients’ decisions in view of their child wish. Hum Reprod.

2011 Apr 1;26(4):861–72.

95. Provoost V, Pennings G, De Sutter P, Dhont M. Decisions on embryo disposition in cross-border

reproductive care: differences between Belgian and Dutch patients at a Belgian fertility center.

Facts Views Vis ObGyn. 2011;3(4):293–301.

58

96. Provoost V, Pennings G, Sutter PD, Velde AV de, Dhont M. Trends in embryo disposition

decisions: patients’ responses to a 15-year mailing program. Hum Reprod. 2012 Feb 1;27(2):506–

14.

97. Lyerly AD, Steinhauser K, Voils C, Namey E, Alexander C, Bankowski B, et al. Fertility patients’

views about frozen embryo disposition: results of a multi-institutional U.S. survey. Fertil Steril.

2010 Jan 15;93(2):499–509.

98. Lyerly AD, Nakagawa S, Kuppermann M. Decisional conflict and the disposition of frozen

embryos: implications for informed consent. Hum Reprod. 2011 Mar 1;26(3):646–54.

99. Cattapan A, Doyle A. Patient Decision-Making About the Disposition of Surplus Cryopreserved

Embryos in Canada. J Obstet Gynaecol Can. 2016 Jan;38(1):60–6.

100. Personal communication with multiple reproductive endocrinology and infertility specialists,

embryologists and fertility nurses, 2016. Toronto, Ontario.

101. Bouwmans CAM, Lintsen BAME, Al M, Verhaak CM, Eijkemans RJC, Habbema JDF, et al.

Absence from work and emotional stress in women undergoing IVF or ICSI: An analysis of IVF-

related absence from work in women and the contribution of general and emotional factors. Acta

Obstet Gynecol Scand. 2008 Jan 1;87(11):1169–75.

102. Kjellberg AT, Carlsson P, Bergh C. Randomized single versus double embryo transfer: obstetric

and paediatric outcome and a cost-effectiveness analysis. Hum Reprod. 2006 Jan 1;21(1):210–6.

103. Fiddelers AAA, Montfoort APA van, Dirksen CD, Dumoulin JCM, Land JA, Dunselman GAJ, et

al. Single versus double embryo transfer: cost-effectiveness analysis alongside a randomized

clinical trial. Hum Reprod. 2006 Aug 1;21(8):2090–7.

104. Kelly J, Hughes CM, Harrison RF. The hidden costs of IVF. Ir Med J. 2006 May;99(5):142–3.

105. Little SE, Ratcliffe J, Caughey AB. Cost of Transferring One Through Five Embryos Per In Vitro

Fertilization Cycle From Various Payor Perspectives: Obstet Gynecol. 2006 Sep;108(3, Part

1):593–601.

106. Petrou S, Trinder J, Brocklehurst P, Smith L. Economic evaluation of alternative management

methods of first-trimester miscarriage based on results from the MIST trial. BJOG Int J Obstet

Gynaecol. 2006 Aug 1;113(8):879–89.

107. Statistics Canada Government of Canada. Table 282-0151 - Labour force survey estimates (LFS),

wages of employees by type of work, National Occupational Classification (NOC), sex, and age

group, unadjusted for seasonality [Internet]. 2016 [cited 2016 May 10]. Available from:

http://www5.statcan.gc.ca/cansim/a26?lang=eng&retrLang=eng&id=2820151&&pattern=&stByV

al=1&p1=1&p2=38&tabMode=dataTable&csid=

108. Ontario Case Cost Initiative Costing Analysis Tool [Internet]. [cited 2016 Jul 14]. Available from:

https://hsimi.ca/occp/occpreports/

109. Statistics Canada, Government of Canada. Table 326-0021 - Consumer price index, annual

(2002=100 unless otherwise noted) [Internet]. Statistics Canada. 2016 [cited 2016 Apr 22].

Available from: http://www5.statcan.gc.ca/cansim/a26?lang=eng&id=3260021

59

110. Ministry of Health & Long Term Care. Schedule of Benefits, Physician Services Under the Health

Insurance Act [Internet]. 2016 [cited 2016 Mar 10]. Available from:

http://www.health.gov.on.ca/english/providers/program/ohip/sob/physserv/sob_master20160229.p

df

111. CARTR/BORN Data Set.

112. Van Den Eeden SK, Shan J, Bruce C, Glasser M. Ectopic pregnancy rate and treatment utilization

in a large managed care organization. Obstet Gynecol. 2005;105(5, Part 1):1052–1057.

113. Hoover KW, Tao G, Kent CK. Trends in the Diagnosis and Treatment of Ectopic Pregnancy in the

United States: Obstet Gynecol. 2010 Mar;115(3):495–502.

114. Trabert B, Holt VL, Yu O, Van Den Eeden SK, Scholes D. Population-Based Ectopic Pregnancy

Trends, 1993–2007. Am J Prev Med. 2011 May;40(5):556–60.

115. Sharma V, Allgar V, Rajkhowa M. Factors influencing the cumulative conception rate and

discontinuation of in vitro fertilization treatment for infertility. Fertil Steril. 2002 Jul;78(1):40–6.

116. Shapiro BS, Daneshmand ST, Restrepo H, Garner FC, Aguirre M, Hudson C. Matched-cohort

comparison of single-embryo transfers in fresh and frozen-thawed embryo transfer cycles. Fertil

Steril. 2013;99(2):389–392.

117. Fiorentino F, Biricik A, Bono S, Spizzichino L, Cotroneo E, Cottone G, et al. Development and

validation of a next-generation sequencing–based protocol for 24-chromosome aneuploidy

screening of embryos. Fertil Steril. 2014 May;101(5):1375–1382.e2.

118. Grifo J, Colls P, Ribustello L, Escudero T, Liu E, Munne S. Why do array-CGH (ACGH) euploid

embryos miscarry? Reanalysis by NGS reveals undetected abnormalities which would have

prevented 56% of the miscarriages. Fertil Steril. 2015 Sep 1;104(3):e14.

119. Tortoriello DV, Dayal M, Beyhan Z, Yakut T, Keskintepe L. Reanalysis of human blastocysts with

different molecular genetic screening platforms reveals significant discordance in ploidy status. J

Assist Reprod Genet. 2016 Nov 1;33(11):1467–71.

120. Tiegs A, Hodes-Wertz B, McCulloh D, Munné S, Grifo J. Discrepant diagnosis rate of array

comparative genomic hybridization in thawed euploid blastocysts. J Assist Reprod Genet.

2016;33(7):893–7.

121. Briggs AH, Goeree R, Blackhouse G, O’Brien BJ. Probabilistic Analysis of Cost-Effectiveness

Models: Choosing between Treatment Strategies for Gastroesophageal Reflux Disease. Med Decis

Making. 2002 Aug 1;22(4):290–308.

122. Fee Schedules from Ontario Fertility Clinics: Anova Fertility, Astra Fertility Group, Hannam

Fertility Centre, IVF Canada, KARMA, London Health Sciences Centre, Markham Fertility

Centre, Mount Sinai Fertility, Nahal Fertility Program, NewLife Fertility Centre, One Fertility,

Ottawa Fertility Centre, Procrea Fertility Centre, Reproductive Care Centre, RepreMed, TRIO

Fertility.

123. Geraedts J, Sermon K. Preimplantation genetic screening 2.0: the theory. MHR Basic Sci Reprod

Med. 2016 Aug 1;22(8):839–44.

60

124. Gleicher N, Kushnir VA, Barad DH. The impact of patient preselection on reported IVF outcomes.

J Assist Reprod Genet. 2016 Apr 1;33(4):455–9.

125. Center for Disease Control and Prevention. 2014 Assisted Reproductive Technology Fertility

Clinic Success Rates Report (National Summary Table) [Internet]. [cited 2017 Mar 14]. Available

from:

https://ftp.cdc.gov/pub/Publications/art/Clinic_PDFs/2014/ART_9999_2014_Fertility_Clinic_Rep

ort.pdf

126. Better Outcomes Registry & Network Ontario. Canadian Assisted Reproductive Technologies

Register (CARTR) Plus: Preliminary treatment cycle data for 2014 & final treatment cycle and

pregnancy outcome data for 2013. [Internet]. CFAS 2015; 2015 Oct [cited 2016 Jan 21]; Halifax,

NS. Available from: http://www.cfas.ca/images/stories/BORN_CARTR_Plus_presentation.pdf

127. Schufreider A, McQueen D, Mathews J, Liebermann J, Uhler ML, Feinberg EC. Is

preimplantation genetic screening with frozen single embryo transfer superior to fresh in-vitro

fertilization with elective single embryo transfer in a good prognosis population? Fertil Steril.

2016 Sep 1;106(3):e60.

128. Forman E, Hong K, Ferry K, Tao X, Taylor D, Levy B, et al. In vitro fertilization with single

euploid blastocyst transfer: a randomized controlled trial. Fertil Steril. 2013 Jul;100(1):100–7.

129. Gleicher N, Barad DH. A review of, and commentary on, the ongoing second clinical introduction

of preimplantation genetic screening (PGS) to routine IVF practice. J Assist Reprod Genet. 2012

Nov 1;29(11):1159–66.

130. Orvieto R. Preimplantation genetic screening- the required RCT that has not yet been carried out.

Reprod Biol Endocrinol. 2016;14:35.

131. Maxwell SM, Colls P, Hodes-Wertz B, McCulloh DH, McCaffrey C, Wells D, et al. Why do

euploid embryos miscarry? A case-control study comparing the rate of aneuploidy within

presumed euploid embryos that resulted in miscarriage or live birth using next-generation

sequencing. Fertil Steril. 2016 Nov 1;106(6):1414–1419.e5.

132. Scott Jr. RT, Galliano D. The challenge of embryonic mosaicism in preimplantation genetic

screening. Fertil Steril. 2016 May; 105(5):1150-2.

133. Arts DGT, Keizer NF de, Scheffer G-J. Defining and Improving Data Quality in Medical

Registries: A Literature Review, Case Study, and Generic Framework. J Am Med Inform Assoc.

2002 Nov 1;9(6):600–11.

134. BORN Ontario. History of BORN Ontario [Internet]. [cited 2016 Aug 16]. Available from:

https://www.bornontario.ca/assets/documents/History%20of%20BORN%20Ontario_May%20201

4.pdf

135. About BORN [Internet]. BORN Ontario. [cited 2016 Aug 16]. Available from:

https://www.bornontario.ca/en/about-born/

136. Gunby J. Assisted reproductive technologies (ART) in Canada: 2012 results from the Canadian

ART register. [Internet]. Montreal, Quebec: Canadian Fertility and Andrology Society; 2014 [cited

2014 Oct 22]. Available from: http://www.cfas.ca/images/stories/pdf/CARTR_2012.pdf

61

137. 2016 Press Release - Results from the Canadian ART Register [Internet]. Canadian Fertility &

Andrology Society. [cited 2017 Apr 14]. Available from: https://cfas.ca/public-affairs/canadian-

art-register/

62

7 Tables

7.1 Description of clinical inputs included in the model.

Age <

35

Age 35

- 37

Age 38

- 40

Age 41

- 42

Age >

42

Minimum Maximum Distribution Source

Pregnancy rate after a euploid

embryo transfer 0.521 0.539 0.510 0.516 0.444 -15% +15% Triangle (79)

Pregnancy loss rate after IVF

with PGS 0.119 0.113 0.131 0.133 0.182 -15% +15% Triangle (79)

Pregnancy rate after an

unscreened embryo transfer 0.486 0.424 0.326 0.222 0.115 -15% +15% Triangle (79)

Pregnancy loss rate after IVF

alone 0.112 0.139 0.200 0.285 0.355 -15% +15% Triangle (79)

Drop out after failed transfer 0.056 0.023 0.090 Triangle (83–90)

Drop out after pregnancy loss 0.066 0.027 0.106 Triangle (90)

63

7.2 Detailed description of all costs included in the model.

Mean Minimum Maximum Standard Deviation Distribution Source

Fertility Treatment

ICSI $1,600.00 $1,500.00 $2,000.00 $200.00 Gamma (122)

Embryo biopsy and PGS $3,961.02 $3,058.07 $5,100.00 $828.99 Gamma (122)

Cryopreservation and 1 year of storage $971.00 $850.00 $1,230.00 $139.15 Gamma (122)

Frozen embryo transfer cycle $1,975.71 $1,500.00 $2,480.00 $315.95 Gamma (122)

Cost of COH and OPU +/- fresh ET $7,985.71 $6,000.00 $11,000.00 $1,495.16 Gamma (122)

Cost of embryo storage for an additional

year $440.00 $200.00 $500.00 $88.80 Gamma (122)

Medications

IVF cycle $4,077.55 $2,718.37 $5,436.74 $1,109.77 Gamma (100,111)

Frozen embryo transfer cycle $184.63 $123.09 $246.18 $50.25 Gamma (100)

Luteal support after fresh embryo transfer $140.28 $93.52 $187.04 $38.18 Gamma (100)

Luteal support after frozen embryo

transfer $465.19 $310.13 $620.26 $126.61 Gamma (100)

Prenatal Care, Labour & Delivery

Prenatal consult $77.20 $77.20 $77.20 $0.00 n/a (110)

Prenatal visit $33.70 $33.70 $33.70 $0.00 n/a (110)

Early ultrasound $59.85 $59.85 $59.85 $0.00 n/a (110)

Anatomy ultrasound $75.30 $75.30 $75.30 $0.00 n/a (110)

Prenatal screening $183.00 $183.00 $183.00 $0.00 n/a (110)

64

Spontaneous vaginal delivery (obstetrics) $498.70 $498.70 $498.70 $0.00 n/a (110)

Operative vaginal delivery (obstetrics) $535.60 $535.60 $535.60 $0.00 n/a (110)

C-Section (obstetrics) $579.80 $579.80 $579.80 $0.00 n/a (110)

Operative vaginal delivery (anesthesia) $150.10 $135.09 $210.14 $32.43 Gamma (110)

C-Section (anesthesia) $225.15 $165.11 $315.21 $61.69 Gamma (110)

Spontaneous vaginal delivery (hospital) $2,991.18 $64.96 $77,100.67 $2,105.45 Gamma (108)

Operative vaginal delivery (hospital) $3,848.24 $229.50 $43,239.87 $2,410.06 Gamma (108)

C-Section (hospital) $4,538.24 $356.41 $294,479.69 $5,353.91 Gamma (108)

Newborn care $4,032.87 -25% +25% - Triangle (108)

Pregnancy Loss

Medical management (gynecology) $161.15 $161.15 $161.15 $0.00 n/a (110)

Surgical management (gynecology) $112.40 $112.40 $112.40 $0.00 n/a (110)

Medical management of ectopic

pregnancy (gynecology) $207.80 $207.80 $207.80 $0.00 n/a (110)

Surgical treatment of ectopic pregnancy

(gynecology) $306.85 $306.85 $306.85 $0.00 n/a (110)

Follow up visit after medical

management $33.70 $33.70 $33.70 $0.00 n/a (110)

Surgical management (anesthesia) $135.09 $120.08 $150.10 $12.26 Gamma (110)

Surgical treatment of ectopic pregnancy

(anesthesia) $225.15 $195.13 $315.21 $51.02 Gamma (110)

Surgical management (hospital) $3,296.16 $689.20 $21,764.97 $2,155.42 Gamma (108)

Medical management of ectopic

pregnancy (hospital) $1,842.34 $235.10 $12,681.50 $1,733.65 Gamma (108)

65

Surgical treatment of ectopic pregnancy

(hospital) $4,082.32 $539.55 $24,804.43 $1,883.43 Gamma (108)

Productivity

IVF cycle excluding embryo transfer $386.55 $257.70 $515.40 $105.21 Gamma (100)

Fresh embryo transfer $103.08 $68.72 $137.44 $28.05 Gamma (100)

Frozen embryo transfer $128.85 $85.90 $171.80 $35.07 Gamma (100)

First trimester care $154.62 $103.08 $206.16 $42.08 Gamma (100)

Second and third trimester care $360.78 $240.52 $481.04 $98.19 Gamma (100)

Pregnancy loss - expectant management $1,931.72 $1,287.81 $2,575.63 $525.75 Gamma (106)

Pregnancy loss - medical management $2,191.48 $1,460.99 $2,921.97 $596.45 Gamma (106)

Pregnancy loss - surgical management $2,032.74 $1,355.16 $2,710.32 $553.24 Gamma (106)

Ectopic pregnancy $3,092.40 $2,061.60 $4,123.20 $841.64 Gamma (100)

Partner Productivity

IVF cycle excluding embryo transfer $298.50 $199.00 $398.00 $81.24 Gamma (100)

Fresh embryo transfer $119.40 $79.60 $159.20 $32.50 Gamma (100)

Frozen embryo transfer $119.40 $79.60 $159.20 $32.50 Gamma (114)

66

7.3 Costs included under the four different perspectives examined.

Perspective

Societal Direct Health

Care

Hypothetical

MOHLTC

Patient Self-

pay

Fertility Treatment

ICSI

Embryo biopsy and PGS

Cryopreservation and 1 year of storage

Frozen embryo transfer cycle

Cost of COH and OPU +/- fresh ET

Cost of embryo storage for an additional year

Medications

IVF cycle

Frozen embryo transfer cycle

Luteal support after fresh embryo transfer

Luteal support after frozen embryo transfer

Prenatal Care, Labour & Delivery

Prenatal consult

Prenatal visit

Early ultrasound

Anatomy ultrasound

Prenatal screening

Spontaneous vaginal delivery (obstetrics)

Operative vaginal delivery (obstetrics)

C-Section (obstetrics)

Operative vaginal delivery (anesthesia)

C-Section (anesthesia)

Spontaneous vaginal delivery (hospital)

67

Operative vaginal delivery (hospital)

C-Section (hospital)

Newborn care

Pregnancy Loss

Medical management (gynecology)

Surgical management (gynecology)

Medical management of ectopic pregnancy

(gynecology)

Surgical treatment of ectopic pregnancy (gynecology)

Follow up visit after medical management

Surgical management (anesthesia)

Surgical treatment of ectopic pregnancy (anesthesia)

Surgical management (hospital)

Medical management of ectopic pregnancy (hospital)

Surgical treatment of ectopic pregnancy (hospital)

Productivity

IVF cycle excluding embryo transfer

Fresh embryo transfer

Frozen embryo transfer

First trimester care

Second and third trimester care

Pregnancy loss - expectant management

Pregnancy loss - medical management

Pregnancy loss - surgical management

Ectopic pregnancy

Partner Productivity

IVF cycle excluding embryo transfer

Fresh embryo transfer

Frozen embryo transfer

68

7.4 Point estimates for C, E and ICER by age category and number of blastocysts (societal perspective).

< 35 35-37 38-40 41-42 >42

1

C 5881.030 6306.930 5840.490 5631.990 5437.020

E -0.116 -0.057 -0.052 -0.024 -0.014

ICER -50596.320 -111245.150 -112193.070 -233973.630 -391571.290

2

C 4927.600 5180.290 4081.990 3326.520 2675.850

E -0.138 -0.068 -0.072 -0.035 -0.022

ICER -35760.100 -76265.200 -56841.880 -95687.780 -119556.060

3

C 5141.640 5088.110 3531.280 2222.260 1063.750

E -0.124 -0.059 -0.073 -0.036 -0.026

ICER -41612.340 -85688.230 -48538.180 -61633.660 -40274.750

4

C 5521.190 5190.430 3305.190 1443.430 -289.360

E -0.098 -0.044 -0.063 -0.031 -0.027

ICER -56366.290 -117970.850 -52092.630 -46673.990 10810.060

5

C 5907.410 5357.600 3270.060 908.750 -1419.930

E -0.072 -0.028 -0.048 -0.022 -0.024

ICER -81482.040 -192118.010 -66420.590 -42095.040 58860.770

6

C 6216.200 5495.040 3268.240 417.480 -2583.150

E -0.051 -0.014 -0.033 -0.010 -0.019

ICER -121698.940 -402997.950 -98075.830 -43162.870 135563.900

7

C 6472.570 5645.450 3391.210 198.400 -3359.110

E -0.034 -0.002 -0.018 0.004 -0.012

ICER -187952.860 -2925842.820 -192933.490 54513.310 278717.000

8

C 6663.330 5767.680 3524.300 76.760 -3995.120

E -0.022 0.007 -0.003 0.018 -0.004

ICER -301578.890 806008.300 -1183506.860 4385.750 1131496.230

9

C 6800.970 5862.500 3678.320 24.900 -4511.670

E -0.013 0.014 0.010 0.031 0.006

ICER -514876.490 418591.820 368194.220 794.800 -732773.130

10

C 6556.700 5754.010 3805.350 21.870 -4926.470

E -0.029 0.007 0.021 0.045 0.017

ICER -227263.150 784384.270 183719.870 489.180 -294738.490 The table has been colour coded such that red = northwest quadrant, blue = northeast quadrant, yellow = southwest quadrant and

green = southeast quadrant. Northwest quadrant – IVF/PGS more expensive and less effective than IVF alone; Northeast

quadrant – IVF/PGS is more expensive and more effective than IVF alone; Southwest quadrant – IVF/PGS is less expensive and

less effective than IVF alone; Southeast quadrant – IVF/PGS is less expensive and more effective than IVF alone.

69

7.5 Results of the probabilistic sensitivity analysis by age category and number of blastocysts (societal perspective).

Age < 35 1 2 3 4 5 6 7 8 9 10

NW 99.10% 99.50% 98.90% 99.60% 99.30% 98.20% 96.10% 89.10% 81.90% 98.40%

NE 0.00% 0.00% 0.00% 0.00% 0.30% 1.40% 3.60% 10.80% 18.00% 1.50%

SW 0.90% 0.50% 1.10% 0.40% 0.40% 0.40% 0.30% 0.10% 0.10% 0.10%

SE 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00%

Age 35 – 37 1 2 3 4 5 6 7 8 9 10

NW 97.30% 95.50% 91.40% 87.70% 81.00% 68.50% 52.70% 36.00% 20.30% 35.20%

NE 2.20% 4.00% 8.10% 12.10% 18.50% 31.20% 47.30% 64.00% 79.70% 64.70%

SW 0.50% 0.50% 0.50% 0.20% 0.50% 0.30% 0.00% 0.00% 0.00% 0.10%

SE 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00%

Age 37 – 40 1 2 3 4 5 6 7 8 9 10

NW 98.80% 98.20% 96.90% 94.20% 91.10% 82.90% 71.00% 53.90% 36.80% 25.00%

NE 0.70% 0.80% 2.60% 4.90% 8.50% 16.90% 28.80% 45.90% 63.10% 75.00%

SW 0.50% 1.00% 0.50% 0.90% 0.40% 0.20% 0.20% 0.20% 0.10% 0.00%

SE 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00%

Age 41 – 42 1 2 3 4 5 6 7 8 9 10

NW 96.00% 92.80% 87.90% 77.70% 56.20% 33.80% 22.70% 13.20% 7.90% 4.60%

NE 3.40% 6.40% 11.60% 16.60% 25.00% 30.30% 34.90% 39.20% 41.70% 46.40%

SW 0.60% 0.80% 0.50% 5.60% 15.90% 25.30% 25.20% 18.10% 13.80% 8.50%

SE 0.00% 0.00% 0.00% 0.10% 2.90% 10.60% 17.20% 29.50% 36.60% 40.50%

Age > 42 1 2 3 4 5 6 7 8 9 10

NW 98.10% 96.80% 81.10% 29.70% 7.60% 0.80% 0.30% 0.00% 0.00% 0.00%

NE 1.30% 3.10% 6.20% 4.50% 2.70% 0.50% 1.00% 0.40% 0.10% 0.50%

SW 0.60% 0.10% 12.50% 61.40% 78.20% 74.70% 66.10% 55.00% 42.00% 35.50%

SE 0.00% 0.00% 0.20% 4.40% 11.50% 24.00% 32.60% 44.60% 57.90% 64.00% All numbers are percentages of PSA trial iterations that fall into a given quadrant of the cost-effectiveness plane. NW (northwest

quadrant) – IVF/PGS more expensive and less effective than IVF alone; NE (northeast quadrant) – IVF/PGS is more expensive

and more effective than IVF alone; SW (southwest quadrant) – IVF/PGS is less expensive and less effective than IVF alone; SE

(southeast quadrant) – IVF/PGS is less expensive and more effective than IVF alone. The table has been colour coded such that

red = northwest quadrant, blue = northeast quadrant, yellow = southwest quadrant and green = southeast quadrant.

70

7.6 Point estimates for C, E and ICER by age category and number of blastocysts (alternative perspectives).

(a) Direct Health Care Perspective # Blasts < 35 35-37 38-40 41-42 >42

1

DC 6019.96 6443.49 6050.08 588968 5688.18

DE -0.116 -0.057 -0.052 -0.024 -0.014

ICER -51791.63 -113653.79 -116219.22 -244679.25 -409660.04

2

DC 5119.95 5383.91 4432.15 3797.16 3162.69

DE -0.138 -0.068 -0.072 -0.035 -0.022

ICER -37005.84 -79262.89 -61717.9 -109225.93 -141307.91

3

DC 5295.51 5311.62 3961.19 2851.63 1750.65

DE -0.124 -0.059 -0.073 -0.036 -0.026

ICER -42857.7 -89452.42 -54447.47 -79089.06 -66281.13

4

DC 5641.03 5412.24 3776.55 2189.98 566.85

DE -0.098 -0.044 -0.063 -0.031 -0.027

ICER -57589.69 -123012.31 -59521.64 -70814.17 -21176.38

5

DC 5992.98 5569.24 3759.32 1740.79 -420.88

DE -0.072 -0.028 -0.049 -0.022 -0.024

ICER -82662.42 -199707.19 -76358.37 -80636.34 17446.77

6

DC 6272.8 5694.43 3761.46 1310.92 -1464.06

DE -0.051 -0.014 -0.033 -0.01 -0.019

ICER -122806.92 -418657.58 -112876.72 -135536.34 76833.77

7

DC 6506.58 5833.38 3880.57 1135.09 -2139.57

DE -0.034 -0.002 -0.018 0.004 -0.012

ICER -188940.57 -3023239.06 -220774.26 311881.05 177527.34

8

DC 6680.57 5945.96 4015.73 1043.06 -2691.96

DE -0.022 0.007 -0.003 0.018 -0.004

ICER -302359.06 830922.12 -1344759.74 59596.55 762413.96

9

DC 6806.11 6033.13 4150.41 1010.68 -3139.3

DE -0.013 0.014 0.01 0.031 0.006

ICER -515266.05 430775.35 415450.13 32263.66 509876.48

10

DC 6583.22 5934.35 4268.48 1019.67 -3497.23

DE -0.029 0.007 0.021 0.045 0.017

ICER -228182.53 808968.18 206079.48 22808.93 -209230.92 The table has been colour coded such that red = northwest quadrant, blue = northeast quadrant, yellow = southwest quadrant and

green = southeast quadrant. Northwest quadrant – IVF/PGS more expensive and less effective than IVF alone; Northeast

quadrant – IVF/PGS is more expensive and more effective than IVF alone; Southwest quadrant – IVF/PGS is less expensive and

less effective than IVF alone; Southeast quadrant – IVF/PGS is less expensive and more effective than IVF alone.

71

(b) Hypothetical MOHLTC Perspective # Blasts < 35 35-37 38-40 41-42 >42

1

DC 2922.48 3432.82 3451.55 3683.31 3796.410

DE -0.116 -0.057 -0.052 -0.024 -0.014

ICER -25143.03 -60550.02 -66302.65 -153018.37 -273415.670

2

DC 2726.92 3320.43 3245.39 3538.75 3685.770

DE -0.138 -0.068 -0.072 -0.035 -0.022

ICER -19709.52 -48883.92 -45192.22 -101792.47 -164678.650

3

DC 2861.600 3390.170 3217.310 3488.890 3620.050

DE -0.124 -0.059 -0.073 -0.036 -0.026

ICER -23159.550 -57093.500 -44222.630 -96763.240 -137058.770

4

DC 3092.670 3524.060 3287.480 3504.830 3591.540

DE -0.098 -0.044 -0.063 -0.031 -0.027

ICER -31573.270 -80096.590 -51813.530 -113330.470 -134172.900

5

DC 3321.900 3666.160 3406.050 3564.960 3593.590

DE -0.072 -0.028 -0.049 -0.022 -0.024

ICER -45819.560 -131464.580 -69182.770 -165135.020 -148966.250

6

DC 3514.670 3792.870 3542.770 3653.300 3620.550

DE -0.051 -0.014 -0.033 -0.010 -0.019

ICER -68809.040 -278853.750 -106313.990 -377714.660 -190006.400

7

DC 3664.360 3896.730 3680.030 3758.170 3667.590

DE -0.034 -0.002 -0.018 0.004 -0.012

ICER -106406.970 -2019542.560 -209364.930 1032609.280 -304312.780

8

DC 3775.350 3977.740 3808.280 3871.130 3730.620

DE -0.022 0.007 -0.003 0.018 -0.004

ICER -170870.460 555871.220 -1275290.540 221182.210 -1056584.930

9

DC 3855.250 4038.890 3922.960 3986.230 3806.160

DE -0.013 0.014 0.010 0.031 0.006

ICER -291866.810 288383.100 392683.260 127251.730 618185.820

10

DC 3714.700 3978.840 4018.020 4099.280 3891.260

DE -0.029 0.007 0.021 0.045 0.017

ICER -128756.100 542394.190 193987.520 91696.450 232804.540 The table has been colour coded such that red = northwest quadrant, blue = northeast quadrant, yellow = southwest quadrant and

green = southeast quadrant. Northwest quadrant – IVF/PGS more expensive and less effective than IVF alone; Northeast

quadrant – IVF/PGS is more expensive and more effective than IVF alone; Southwest quadrant – IVF/PGS is less expensive and

less effective than IVF alone; Southeast quadrant – IVF/PGS is less expensive and more effective than IVF alone.

72

(c) Patient Self-Pay Perspective # Blasts < 35 35-37 38-40 41-42 >42

1

DC 7058.500 6971.690 6559.550 6167.390 5852.790

DE -0.116 -0.057 -0.052 -0.024 -0.014

ICER -60726.490 -122970.410 -126005.880 -256216.380 -421515.060

2

DC 6354.060 6024.500 5147.780 4219.440 3427.950

DE -0.138 -0.068 -0.072 -0.035 -0.022

ICER -45925.660 -88693.830 -71683.070 -121372.710 -153606.140

3

DC 6394.930 5882.470 4704.900 3323.760 2091.610

DE -0.124 -0.059 -0.073 -0.036 -0.026

ICER -51755.500 -99066.020 -64669.920 -92183.400 -79190.400

4

DC 6509.380 5849.210 4450.090 2646.170 936.330

DE -0.098 -0.044 -0.063 -0.031 -0.027

ICER -66454.800 -132943.810 -70137.210 -85565.440 -34979.580

5

DC 6632.110 5864.000 4314.290 2136.850 -53.440

DE -0.072 -0.028 -0.049 -0.022 -0.024

ICER -91478.000 -210280.730 -87630.780 -98982.730 2215.420

6

DC 6719.150 5862.580 4179.710 1618.640 -1123.580

DE -0.051 -0.014 -0.033 -0.010 -0.019

ICER -131545.520 -431019.930 -125428.010 -167351.490 58965.750

7

DC 6803.250 5897.670 4161.560 1337.040 -1846.140

DE -0.034 -0.002 -0.018 0.004 -0.012

ICER -197555.150 -3056556.770 -236760.430 367617.290 153180.780

8

DC 6866.240 5929.250 4168.470 1132.950 -2461.560

DE -0.022 0.007 -0.003 0.018 -0.004

ICER -310762.320 828585.960 -1395907.310 64732.420 697161.650

9

DC 6911.880 5955.260 4188.460 985.470 -2984.440

DE -0.013 0.014 0.010 0.031 0.006

ICER -523273.730 425215.380 419259.390 31459.040 -484724.690

10

DC 6829.540 5916.530 4211.480 881.410 -3427.470

DE -0.029 0.007 0.021 0.045 0.017

ICER -236720.220 806538.990 203327.560 19716.120 -205057.290 The table has been colour coded such that red = northwest quadrant, blue = northeast quadrant, yellow = southwest quadrant and

green = southeast quadrant. Northwest quadrant – IVF/PGS more expensive and less effective than IVF alone; Northeast

quadrant – IVF/PGS is more expensive and more effective than IVF alone; Southwest quadrant – IVF/PGS is less expensive and

less effective than IVF alone; Southeast quadrant – IVF/PGS is less expensive and more effective than IVF alone.

73

7.7 Causes of inadequate PGS data quality and suggestions for data quality improvement in the CARTR Plus / BORN Ontario registry.

Likely Cause of Data Quality Issue Proposed Action to Improve Data Quality

Registry set up & organization

- PGS inappropriately captured under

“Reason for Treatment Cycle”

- Divide “Reason for Treatment Cycle” into “Fertility

Diagnoses” and “Other Reasons for ART”

- Remove PGS from “Reason for Treatment Cycle”

and create new section for capturing data related to

PGS

- PGS results not captured

- Create new section for capturing data related to

PGS including whether or not PGS was planned

and/or ultimately performed

- Include precise genetic results for each embryo

screened

- Embryo stage at biopsy and PGS

platform not captured

- Include data points for embryo stage at biopsy and

PGS platform used in new PGS section

Data collection

- Misclassification of PGS and PGD

cases, particularly for cases with

balanced translocations or other

structural abnormalities

- For all cases where PGS and/or PGD is planned,

include a data point for the diagnosis of interest (or

“routine aneuploidy screening”)

- Educate end users regarding the correct way to

input these cases

- Inconsistent handling of cases where

PGS was planned but not completed

- Include data points to capture data related to

whether or not PGS was planned and/or ultimately

performed

- Include data points for reasons why PGS was not

performed if it was planned

- Educate end users regarding the importance of

accurate data entry

Quality improvement

- Lack of feedback mechanism for

end users of the data back to registry

organizers

- Create an easy and efficient mechanism for end

users and researchers to provide feedback to

CARTR Plus / BORN Ontario on an ongoing basis

- Regularly update and improve the database,

particularly as it relates to rapidly evolving

technologies

74

8 Figures

8.1 The cost-effectiveness plane.

From Hoch JS, Briggs AH, Willan AR. Something old, something new, something borrowed, something blue: a

framework for the marriage of health econometrics and cost‐effectiveness analysis. Health Econ. 2002;11(5):415–

30. (25)

75

8.2 Simplified diagram of the Markov model used within the IVF alone arm of the full decision tree.

The purple circle represents the Markov node.

76

8.3 Simplified diagram of the IVF/PGS arm of the decision model, including the embedded Markov model and its clones.

The purple circle represents the Markov node. The purple ovals with the letter M in the centre of them represent

clones of the illustrated Markov process.

77

8.4 Full decision tree comparing IVF/PGS to IVF alone.

78

8.5 One-way sensitivity analysis: tornado diagrams by age group and number of blastocysts, from a societal perspective.

79

All variables except those costs with known static values were tested in one-way sensitivity analyses; however, only

the variables that had the greatest impact on the resultant ICERs are included in the above tornado diagrams. The

vertical line represents the value of the ICER for that age group and number of blastocysts (see Table 7.4). IR =

implantation rate; EPL = early pregnancy loss; FET = frozen embryo transfer.

80

8.6 Probabilistic sensitivity analysis: scatter plots by age group, from a societal perspective.

81

8.7 Probabilistic sensitivity analysis: scatter plots by age group, from alternative perspectives.

(a) Hypothetical MOHLTC Perspective

82

(b) Patient Self-Pay Perspective

83

8.8 Components of data usability.

84

9 Appendices

9.1 Categorization of causes of inadequate data quality in medical registries.

From Arts DGT, Keizer NF de, Scheffer G-J. Defining and Improving Data Quality in Medical Registries: A

Literature Review, Case Study, and Generic Framework. J Am Med Inform Assoc. 2002 Nov 1;9(6):600–11. (133)