The data were derived from a modelled study population consisting of subfertile women undergoing their first IVF treatment at the Erasmus MC, University Medical Center Rotterdam, the Netherlands. The data of the RCT were used to model nutrition and lifestyle behaviors. In this RCT, participants were randomly assigned to the intervention or control group. Participants of the intervention group received the complete coaching program and were coached on a maximum of five nutrition and lifestyle behaviors: vegetable, fruit, and folic acid supplement intake, smoking, and alcohol consumption. Participants of the control group only received a diminished version of the program. At several time points, all participants were asked to fill out questionnaires about their nutrition and lifestyle behavior. In this way, change in behavior could be measured. Participants of the RCT started the program at a maximum of 2 months before start of their IVF treatment, and the program lasted for a period of 24 weeks (Figure 1). The study protocol and primary results of the RCT on the improvement of these behaviors have been published elsewhere [6,10]. In brief, participants in the intervention group showed a significantly larger improvement in inadequate behavior compared to the control group [6].

A decision tree model was constructed using Microsoft Excel (version 2010; Microsoft Corporation, Redmond, Washington) to assess the incremental ongoing pregnancies following the first IVF cycle and the costs of the mHealth program as compared to usual care (Figure 2). Ongoing pregnancy was defined as a vital pregnancy at 12 weeks of gestation. Women of subfertile couples who underwent their first IVF treatment in 2015 entered the model (n=793). A second IVF cycle was started if the first cycle did not result in an ongoing pregnancy. Pregnancy outcome following the second IVF cycle was the endpoint of the model. This short-term evaluation should therefore be considered a first indication of cost-effectiveness of the mHealth program.

The usual care scenario reflects usual IVF treatment in the Netherlands. We assumed that all women received an elective single embryo transfer and that pregnancy rates in usual care are 33% for the first IVF cycle and 23% for the second cycle [11]. We furthermore assumed that all women in the intervention scenario were offered the mHealth program (100% coverage). This program was not offered in the usual care scenario (0% coverage). The intervention adherence rate was set at 70%, based on RCT data in which 70% of participants in the intervention group completed the coaching [6].

Analyses were performed from a health care and societal perspective. The health care perspective includes costs related to the mHealth program [12], all costs associated with IVF treatment (eg, laboratory and hospital costs), and other relevant health care costs (eg, general practitioner visits). The societal perspective includes all health care costs plus costs outside the health care sector (eg, costs due to absence at work). The model parameters, including their distributions and sources, are reported in Table 1. Ongoing pregnancy rates after the first and second IVF cycle for the Smarter Pregnancy scenario and the usual care scenario were based on our previous study in the same setting [8] and others [13]. A detailed description of the cost calculations has been provided by Fiddelers et al [9]. All costs were expressed in euros (€) for the reference year 2016 based on the Dutch price index [14].

The primary effect outcome measure was expressed as the number of ongoing pregnancies after two IVF cycles. Incremental cost-effectiveness ratios (ICERs) from health care and societal perspectives were calculated by dividing the difference in costs between the Smarter Pregnancy scenario and the usual care scenario by the difference in the number of ongoing pregnancies in both scenarios. The ICER represents the estimated costs of one additional ongoing pregnancy.

A probabilistic sensitivity analysis was performed to consider the uncertainty surrounding the point estimates of the model input parameters. Probabilistic distributions were assigned to the parameters (Table 1). Thereafter, 1000 model iterations were performed by drawing random values from the distributions assigned to the input parameters. We calculated the average costs and ongoing pregnancies by averaging these 1000 iterations. We performed deterministic sensitivity analyses to investigate the impact of changing several key parameters of the model: the coverage and adherence rate of the mHealth program and the chance of an ongoing pregnancy following the use of this program.

This model-based study, determining the estimates of available data, showed that the use of the mHealth program would result in 86 additional pregnancies and a reduction of €270,000 compared to usual care after two IVF cycles, resulting in an ICER of –€3050 per additional ongoing pregnancy. Sensitivity analyses showed that the use of this mHealth program is cost saving when the ongoing pregnancy rate increases to at least 51% after two cycles of single embryo-transfer IVF treatment.

A strength of our model is the combined use of evidence-based data of the population, clinical effectiveness, compliance, and costs to support decision making. Although model parameters would ideally be based on meta-analyses or larger datasets, these were unavailable. Since the Smarter Pregnancy RCT is ongoing, assumptions regarding ongoing pregnancy rates had to be made based on our previous data. In economic evaluations, a time horizon that is long enough to capture all relevant costs and effects is preferred [19]. Our study was limited to two IVF cycles, which may be relatively short. However, as the endpoint of our study was to assess the incremental ongoing pregnancy rate, other costs and long-term reproductive and health outcomes were not considered.

We evaluated single embryo transfers only, because in the Netherlands, this is the most common IVF strategy. Therefore, costs and ongoing pregnancy rates of other IVF strategies will be different [13].

Several studies have investigated the effectiveness of nutrition and lifestyle interventions preceding fertility treatment. However, most of these studies focus on specific patient groups such as obese or anovulatory women [20,21]. In accordance with our findings, the study by Van Oers et al [22] showed that lifestyle intervention preceding fertility treatment was cost-effective in terms of achieving an ongoing pregnancy within 24 months.

The difference in average societal costs and health care costs was relatively small, indicating that the addition of the non-healthcare costs had no substantial impact on the ICER. Because nutrition and lifestyle interventions in preconception care have relatively low additional budget impact, we expect that the chance that the mHealth program is not cost-effective would be low [23].

Our results show that the mHealth coaching program Smarter Pregnancy is potentially cost saving for subfertile couples preceding their first IVF treatment. Although our results are promising, our model requires further validation based on actual data on ongoing pregnancy rates from the Smarter Pregnancy RCT in order to establish the relative cost-effectiveness of the mHealth program with greater certainty. Implementation of this mHealth program in routine preconception care of subfertile couples should be seriously considered, given the relatively low intervention costs and promising cost-effectiveness estimates.