In any health care system, cost-effective commissioning of health care in order to maximise population health outcomes with the minimum possible resource use, is an important consideration. The UK National Institute for Clinical Excellence (NICE) [1] recommends offer of up to six cycles of Unstimulated Intrauterine Insemination (U-IUI) for couples with unexplained or mild male factor subfertility. Supplementary cost calculation guidance suggested overall savings to the NHS from a "switch" of activity from Stimulated IUI (S-IUI) to U-IUI, which was also recommended [2]. These calculations however were based on 'head-to-head' comparisons between S-IUI and U-IUI and completely ignored the cost implications of the fact that many couples who fail IUI will subsequently require in-vitro fertilisation (IVF) treatment.

Even if some health care interventions may have a relatively low cost per treatment session, they could be less cost-effective compared to alternative treatments with a higher cost per treatment session, but also better clinical effectiveness. It is therefore legitimate to examine critically whether funding for IUI allows better management of current demand, or whether it increases activity and overall associated healthcare expenditure for those couples who are eligible for both treatment strategies [3]. We therefore constructed a mathematical model to evaluate the cost-effectiveness of provision of IUI followed by IVF compared to primary offer of IVF for patients who according to NICE are eligible for both treatment strategies.

The findings suggest that for couples with unexplained or mild male factor subfertility, primary offer of a full IVF cycle is less costly and more cost-effective than primary offer of any IUI modality followed by IVF (for IUI failures). Importantly, the cost of primary IUI (any modality) increases and its cost-effectiveness decreases with increasing number of IUI cycles offered. Between S-IUI and U-IUI, S-IUI followed by IVF was associated with higher cost and lower cost-effectiveness compared with U-IUI followed by IVF. The sensitivity analysis findings indicate that the main finding of inferior cost-effectiveness of the primary IUI, compared with primary IVF, is robust. A previous study on cost impact implications of the 2004 NICE guidance has failed to reach similar conclusions, as it did not accurately simulate real life patient experience and clinical management [1].

In principle, a prospective randomised controlled trial would have been a superior study design to answer this study's research question. However, such a study would have had to entertain certain practical obstacles, including ethical clearance, and may not be forthcoming for many years to come. At the same time resource allocation and service development policy decisions have to be made currently and in the near future. In the absence of RCT-based evidence, mathematical modelling synthesising evidence from various high quality sources is the only rational way to support health care policy decision-making [14].

Published data comparing the cost-effectiveness of IVF versus IUI are scarce. Some studies indicate that IUI may be more cost-effective than IVF in cases of unexplained and moderate male subfertility [13, 15–17]. However, these studies have not accounted for the potential subsequent FET component of IVF strategies, which increases their cost-effectiveness [17]. More importantly, these studies do not take into account the potential for eventual "transfer" of most couples who fail IUI to IVF, a crucial consideration by commissioners of healthcare services when both IUI and IVF are funded from the same budget (as is the case in the UK NHS).

The majority of the literature on IUI effectiveness comes from uncontrolled retrospective case series and cohort designs, could overestimate treatment effectiveness because of selection bias and consideration of clinical pregnancy as opposed to live birth as the study outcome. Also, the number of abandoned cycles is generally excluded from the denominator, leading to inflated estimates of effect [11]. A further problem is that the reported effectiveness of IUI is very variable, ranging from small integer numbers to up to over 30%. This heterogeneity may be due to factors including differences in study populations, ovarian stimulation regimes, the number of inseminations per cycle, the timing of insemination, and methods of sperm preparation [18]. Also, the reported trials involved an exceedingly small fraction of the very great number of patients world-wide undergoing treatment, so generalisations have to be cautious [19]. Given the above limitations of published evidence, we used local IUI activity data as the main source of our data in relation to the effectiveness of IUI, but complemented those with different sets of estimates in sensitivity analysis. The results obtained using local current data could better reflect the likely current UK practice in a "real life" situation, and for an unselected population.

We used a mean LBR value of 7% for S-IUI based on local experience on 676 cycles of S-IUI over 10 years. In most studies the pregnancy rate per cycle of S-IUI is approximately 10% [20–22]. A prospective randomised study from The Netherlands showed that the average LBR of S-IUI of 8.7% per cycle [13]. NICE in its economic model has used results of an RCT from the US [23] reporting cumulative pregnancy rates of 33% for S-IUI for an average of 5.6 cycles. In our model, if the cumulative LBR for five and six cycles of S-IUI is considered, it equates to 30% and 35% respectively, and as such our assumptions are comparable to the order of the estimates used by NICE.

The effectiveness (LBR) of IUI was judged to be 3.5% per treatment cycle -this represents an informed judgement, based on the literature. Many experts would support that this is actually a high estimate, but we preferred to use a "generous" estimate. If the true estimate is lower, any bias introduced from this assumption would have favoured the cost-effectiveness of U-IUI, and would have made IVF less cost effective, i.e. if a lower LBR value for U-IUI was used, the results would have been even more favourable for primary offer of IVF as the preferred treatment strategy.

We used average success rates rather than age- and indication-specific LBRs. Maternal age affects outcomes of IUI, but so does for IVF [1]. Similarly, different subfertility indications are associated with different success rates for IUI, but again the same is true for IVF. Therefore, the lack of stratification (by age or indication) is very unlikely to influences the results for either modality. In any case, the hypothetical cohort of 100 couples can be assumed to have an identical case mix in terms of age and indication stratification, whichever treatment strategy is offered.

The modelling of costs has concentrated on the nominal cost of IVF and IUI. Other essential aspects of the service (e.g. diagnostic activity, incident and follow-up appointments, counselling) have not been calculated, because they were assumed to be already occurring (i.e. requiring no 'new' money) and to be similar to both treatment strategies. The cost associated with complications, such as multiple gestation and birth, were also not included in calculations. Evidence suggests that there is no statistically significant difference in the incidence of ovarian hyperstimulation syndrome (OHSS) and multiple pregnancy rate between S-IUI and IVF [13]. As such, it is unlikely that inclusion of the costs of complications would have had an impact on the comparative cost-effectiveness of S-IUI and IVF.

Lastly, anecdotally, undergoing IUI followed by IVF may be more psychologically distressing to the couple compared to IVF alone, as it may be more frustrating to have, on average, comparatively more failed treatment cycles -although direct evidence about this assumption is lacking. However, some patients may have a positive preference for IUI or an aversion for IVF (on philosophical, or religious grounds). We acknowledge that traditional methods of cost-effectiveness analysis like the ones used in this study, are not useful tools when dealing with philosophically-grounded patient preferences – which however are exceptional, rather than the norm.

Our study indicates that for couples with unexplained or mild male factor subfertility, a primary treatment offer of any modality of IUI instead of primary offer of IVF is cost-ineffective, and therefore associated with considerable opportunity costs. The additional and avoidable costs incurred put pressure on the health care system to cope with extra demand and activity for a treatment of low effectiveness, making the wider availability of the most effective treatment (IVF) more difficult, and thus disadvantaging couples who could have otherwise benefited from it.