Does in vitro fertilization (IVF) treatment provide good value for money? A cost-benefit analysis

Background Using traditional health technology assessment (HTA) outcome metrics, such as quality-adjusted life-years, to assess fertility treatments raises considerable methodological challenges because the objective of fertility treatments is to create new life rather than extend, save, or improve health-related quality of life. Objective The aim of this study was to develop a novel cost-benefit framework to assess value for money of publicly funded IVF treatment; to determine the number of cost-beneficial treatment cycles for women of different ages; and to perform an incremental cost-benefit analysis from a taxpayer perspective. Methods We developed a Markov model to determine the net monetary benefit (NMB) of IVF treatment by female age and number of cycles performed. IVF treatment outcomes were monetized using taxpayers' willingness-to-pay values derived from a discrete choice experiment (DCE). Using the current funding environment as the comparator, we performed an incremental analysis of only funding cost-beneficial cycles. Similar outputs to cost-effectiveness analyses were generated, including net-benefit acceptability curves and cost-benefit planes. We created an interactive online app to provide a detailed and transparent presentation of the results. Results The results suggest that at least five publicly funded IVF cycles are cost-beneficial in women aged <42 years. Cost-benefit planes suggest a strong taxpayer preference for restricting funding to cost-beneficial cycles over current funding arrangements in Australia from an economic perspective. Conclusions The provision of fertility treatment is valued highly by taxpayers. This novel cost-benefit method overcomes several challenges of conventional cost-effectiveness methods and provides an exemplar for incorporating DCE results into HTA. The results offer new evidence to inform discussions about treatment funding arrangements.


Comparator:
We conducted an incremental cost-benefit analysis (CBA) using the current funding environment in Australia as the comparator. This allowed us to construct incremental cost-benefit ratios (ICBR) following the approach outlined in McIntosh (1). We defined the comparator as funding the current "package" of fertility treatment in Australia, which is characterized by providing funding for an unlimited number of cycles in all age groups. To facilitate the analysis, we derived the costs and benefits per taxpayer of an average cycle in this package according to Equation 1 below: Using utilization data for 2018 (based on the Australia and New Zealand Assisted Reproduction Database (ANZARD)) to determine the proportion of cycles stratified by age group and complete cycle, we calculated the sum of proportionate costs and benefits in each cycle and age group. In some cases, it was not possible to derive the average costs per taxpayer and treatment cycle from the probabilistic sensitivity analysis output which was due to one of two reasons: 1) The treatment cycle was excluded from the analysis due to a lack of sufficient data on transition probabilities in ANZARD (e.g., 8 th complete cycle in 44-45-year-old women was excluded because available data was based on <30 women).
2) No women started the treatment cycle in the iteration of the Markov model considered due to the model terminating (e.g., if women achieved 4 live births) or due to women dropping out of treatment prior to reaching the treatment cycle considered.
In these cases, we assumed that average costs per taxpayer and cycle were equal to the average costs per taxpayer and cycle from the last complete cycle with an available estimate in the age group considered. For instance, in the example shown in Supplementary Table 1, there was no available cost estimate for the 8 th complete cycle in <30-year-old women. Therefore, we assumed average costs were the same as in the 7 th complete cycle, the last complete cycle with a cost estimate (i.e., $33.11 per taxpayer).
Finally, we summed the proportionate costs (last column in Supplementary Table 1) across all cycles and age groups to obtain the average costs that each taxpayer is funding, on average, for the provision of the current package of fertility treatment. In the example calculation in the table, each taxpayer would contribute a total of $28.42, on average, to fund the current package of fertility treatment.
The calculation for the benefits of an average cycle in the current package is equivalent: Average benefits per taxpayer and complete cycle are based on the model outputs from the probabilistic sensitivity analysis and calculated as the cycle-specific live birth rate multiplied by the willingnessto-pay for a live-born babythe value of a statistical baby. Where such an estimate was not available for a treatment cycle, it was assumed that average benefits per taxpayer are the same as in the last cycle with a benefit estimate. Proportionate benefits are then calculated as the average benefits per taxpayer and cycle multiplied by the proportion of all women in that age group and cycle. Finally, the average benefits per taxpayer for the current package of fertility treatment are derived by summing the proportionate benefits across all cycles and age groups. For the example shown in the table, the average benefits per taxpayer for the current package of fertility treatment were estimated to be $92.08.

Policy intervention:
The 'policy intervention' which we compare to the current funding arrangement in Australia is defined in this study as funding the number of cycles that were deemed cost-beneficial in the CBA (i.e., those that have a positive NMB) by female age. To determine the sum of proportionate costs and benefits in each cycle and age group in this policy intervention package we used the same utilization data (for 2018; based on ANZARD). However, prior to determining the proportion of cycles stratified by age group and complete cycle, we excluded any complete cycles that were not found to provide good value for money. Otherwise, calculations were the same. To continue the example above (Supplementary Table 1), the cost and benefit estimates for the provision of the policy intervention package of fertility treatment in the particular iteration of the probabilistic sensitivity analysis considered were as follows: -The average costs that each taxpayer is funding were $29.22.
-The average benefits that each taxpayer is receiving were $101.33.

Incremental cost-benefit analysis:
In the final step, we determined the incremental cost and the incremental benefit as the difference in average costs and benefits per taxpayer, respectively, between the policy intervention and the current package of fertility treatment (Supplementary Table 2). Wherever, incremental benefits outweigh incremental costs, we would prefer the policy intervention package of fertility treatment. Similarly, we would prefer the current package of fertility treatment where this is not the case. For the example above, incremental costs are $0.80 (i.e., $29.22 -$28.42) and incremental benefits are $9.25 (i.e., $101.33 -$92.08), meaning incremental benefits outweigh incremental costs and, therefore, the policy intervention package of fertility treatment is preferred over the current package of fertility treatment.
Similar to all other analyses (in the main manuscript), we conducted the incremental analyses for two scenarios: 1) We excluded women and their associated costs in each complete cycle who did not reach oocyte pick-up.
2) We considered all women simulated in the model (i.e., 1,000 women per age group in each of the 1,000 iterations of the Markov model in the probabilistic sensitivity analyses) and their associated costs.
In addition to presenting the results of the incremental analysis aggregated for all age groups, we created cost-benefit planes showing the incremental costs and benefits separately for each age group. The respective calculations are equivalent to the aggregated analysis. Supplementary Table 2 includes the (incremental) cost and benefit estimates of the age-stratified analysis corresponding to the example discussed above. The aim of this approach is to show how each age group contributes to the aggregated results, which is why the cost and benefit estimates stratified by age always sum to the aggregated estimates.
For instance, in the example shown in Supplementary Table 2 it can be seen that for women aged <42 years the policy intervention package of fertility treatment provides incremental benefits at additional costs compared to the current package, and that the incremental benefits outweigh the incremental costs. In contrast, for women aged 42-43 years, cost savings in the policy intervention package of fertility treatment are associated with a reduction in benefits. While at least one complete cycle was good value for money and, hence, considered in the policy intervention package, some cycles were not good value and, therefore, excluded. As a result, incremental costs and benefits compared to the current package are negative because the cycles that were excluded were associated with a particular treatment cost and chance of a live birth (i.e., benefit). In Supplementary Table 2 it can also be seen that no cycles in women aged 44-45 years or >45 years in this example were good value, leading to cost and benefit estimates equal to $0 for the policy intervention package. In such cases, incremental costs and benefits would always be negative for the same reason as described above. (2) all women and their costs are considered) we created two cost-benefit planes:

Supplementary
1) A cost-benefit plane for the aggregated analysis where each of the 1,000 iterations of the model in probabilistic sensitivity analyses is represented by one dot (Panel A of Supplementary Figure 1).
2) A cost-benefit plane for the age-stratified analysis where each of the 1,000 iterations of the model in probabilistic sensitivity analyses is represented by 10 dotsone for each age group (Panel B of Supplementary Figure 1).
As a general rule, interventions are considered to be preferred to the comparator if incremental benefits outweigh incremental costs, and we would be indifferent between the intervention and comparator if incremental costs equal incremental benefits. We visualize the points of indifference in the cost-benefit plane by a straight red line through the origin with slope=1. All dots to the right of this red line indicate that the intervention would be preferred to the comparator as incremental benefits are greater than incremental costs.
Supplementary Figure 1. Cost-benefit planes for the example discussed in the appendix.
Panel A: cost-benefit plane for the aggregated analysis; Panel B: cost-benefit plane for the agestratified analysis. The red line indicates points of indifference (i.e., where incremental costs equal incremental benefits).
Limitations: It was not always possible to assign average costs and benefits per taxpayer for a treatment cycle in the current package of fertility treatment (i.e., comparator) based on the probabilistic sensitivity analysis. This was due to one of two reasons: (1) The treatment cycle was

A B
excluded from the analysis due to a lack of sufficient data on transition probabilities in ANZARD (e.g., 8th complete cycle in 44-45-year-old women was excluded because available data was based on <30 women). (2) No women started the treatment cycle in the iteration of the Markov model considered due to the model terminating (e.g., if women achieved 4 live births) or due to women dropping out of treatment prior to reaching the treatment cycle considered. In such cases, we assumed the cost and benefit estimates were equal to the estimates of the last cycle in this age group with such estimates available. Consequently, if that cycle was considered good value for money, the cycle with missing estimates would also be considered good value. However, given that it is not included in the policy intervention package, the cost-benefit plane might show a preference of the current package of fertility treatment in these age groups/in the aggregated analysis over the policy intervention package.

Supplementary
The ICER for publicly funded IVF treatment stratified by complete cycle and female age was derived according to the following equation: We only performed a base-case analysis for the main scenario (i.e., where only women who reached oocyte pick-up (OPU) procedure in the Markov model and their associated costs were considered in each complete cycle) for comparison with our main analysisthe cost-benefit analysis.
While our main goal in performing the cost-effectiveness analysis was to ensure comparability with the CBA results, some modifications in the methodology were necessary that impacted comparisons: (1) We assumed a 20-year time horizon over which QALY improvements following a live birth resulting from IVF treatment were considered and discounted future QALYs at a rate of 3.5% per year. In contrast, discounting was inconsequential for the CBA and benefits resulting from each live birth were realized immediately (valued monetarily using the VSB), rather than accumulated over time. (2) In the CBA, each live birth resulted in the same benefit (equal to the VSB of $223), whereas the literature suggests that QALY utility weights do not increase linearly with the number of children (2). Given limited evidence, we had to make strong assumptions about these utility weights and the preferences of women, which were not required for the CBA.

Results
Supplementary Table 2 reports costs per live birth achieved and costs per QALY gained stratified by female age group and complete cycle and Supplementary Figure 1 provides a visual representation of ICER values against the generally assumed QALY threshold value in Australia of approximately $50,000 (3,4). The results indicate that the number of cycles with good value for money (i.e., with a cost per QALY gained estimate below the $50,000 threshold value) continuously declines with female age with seven cycles providing good value for money in women aged <30 years compared to one cycle in women aged 38-39 years. In women aged 40 years and over no cycles were found to provide good value for money. Notes: The slider at the bottom allows adjusting the limits of the horizontal axis to zoom in/out of the figure. When hovering the cursor above data points in the online figure, data labels appear that provide the exact values for the WTP per taxpayer for a live birth and the corresponding probability that the cycle is good value for money. As an example, the screenshot shows the data label for the 4 th complete cycle in 38-39-year-old women at a WTP value per taxpayer for a live birth of $110.

Supplementary Figure 7. Cost-benefit planes by age group (left panel) and overall (right panel)
showing incremental cost-benefit ratios of only funding cost-beneficial cycles compared to the current funding environment in Australia, which is characterized by unrestricted funding based on age and previous treatment attempts.
Notes: When hovering the cursor above data points in the online figure, data labels appear that provide the exact values for the incremental benefits and incremental costs. As an example, the screenshot shows the data label for one iteration of the model.