Strengths and limitations
To our knowledge, this is the first economic evaluation that considers the cost-effectiveness of early detection of CMD with a stepwise approach. Moreover, this is one of the few large-scale clinical studies that investigated the cost-effectiveness of targeted CMD prevention in primary care alongside a trial to ensure that appropriate outcome and cost data were collected. The second strength is that changes in systolic blood pressure and cholesterol levels were modelled simultaneously, because various CMD risk factors are known to have a multiplicative effect [
34].
However, also some limitations should be addressed.
For the long-term cost-effectiveness analysis, we adopted a healthcare perspective, because the model cannot simulate costs other than healthcare costs up to now. Taking a societal perspective (including broader societal costs and consequences of the intervention such as productivity losses) could have slightly improved the ICER if positive changes in lifestyle had been achieved. Unfortunately, the effectiveness analysis did not show such an effect [
12]. However, benefits of the intervention in terms of prevented CMD events are expected in the future and therefore will not be reflected in productivity costs.
A one-off intervention would not be the “real-world” scenario to implement a preventive intervention. A one-off intervention is most likely more efficient in terms of organization, as fixed costs are distributed over larger groups of patients. In a real-world scenario, the ICER would probably be even higher than our current estimate.
Besides blood pressure and cholesterol levels, no other risk factors were modelled, possibly leading to an underestimation of the reported results. Particularly, the non-inclusion of the non-significant effect that the intervention had on quitting smoking may have contributed to the high ICER that was found. In addition, systolic blood pressure and total cholesterol levels were modelled via discrete classes. The real distribution of these parameters is continuous and even slight improvements in these risk factors will have a favourable outcome on health. Nevertheless, our data from the clinical study also showed substantial numbers of patients with increased levels of blood pressure and cholesterol in the first year after the intervention (Additional file
3). The unfavourable ICERs certainly relate to the fact that the pattern of improvement over patients in the intervention group was not visible in all participating patients. Using discrete levels implies a simplified reflection of the reality; however, the classes used correspond to relative risks based on the best literature available. Furthermore, the model is based on assumptions about the long-term maintenance and changes in risk factors, which could have resulted in a slight under- or overestimation of the long-term outcomes of the intervention. Despite this unavoidable variability, this could not have compromised our main conclusions as only very small health gains were achieved at relatively high costs.
In addition, the model is about 10 years old and the most recent clinical developments in the treatment of diabetes and cardiovascular care have not been incorporated; however, we believe as only small effects were achieved with our intervention these developments could not have been of major influence on our results.
We carefully assessed healthcare costs based on extracted EHR data; however, the EHR contains no data on the use of hospital care. Because the intervention might have led to some hospital referrals, this could have resulted in a slight underestimation of the intervention-related healthcare costs. On the other hand, we might have overestimated the intervention-related patient costs. For example, costs for all new gym subscriptions were considered to be emanating from the intervention. However, we believe that such costs were also made during care as usual.
One final limitation is that the data on patient and family costs and productivity losses were based on self-report. Self-reported data are vulnerable to recall bias. This bias was assumed equal between the intervention and control groups.
Comparison with existing literature and interpretation of results
The advantage of a stepwise screening approach is that only high-risk individuals are targeted—those who are expected to benefit most from preventive treatment—and therefore assumed to be more cost-effective compared to whole population screening [
35]. This is in line with the 2016 guidelines of the European Society of Cardiology which consider targeted CVD screening in high-risk individuals [
4].
Modelling studies have demonstrated that targeted prevention strategies for CVD or DM2 in high-risk individuals are most likely cost-effective [
21,
36‐
40]; however, none of these studies evaluated early detection strategies for CMD (CVD, DM2 and CKD). In addition, most of these studies did not include a control group, which is a risk for overrating economic value, as usual care is associated with better healthcare outcomes than no treatment. Furthermore, simulation modelling studies with a positive cost/benefit ratio generally assume lower intervention costs, higher uptake rates, larger treatment effects and sustained compliance than found in clinical trials [
11]. Nevertheless, economic modelling of clinical trial data remains very important to project results beyond trial duration to estimate its costs and cost-effectiveness, as follow-up in clinical trials in general is too short to observe subsequent disease incidence and mortality.
Despite promising results regarding lifestyle improvement, another large trial investigating the cost-effectiveness of a European prevention program in primary care focusing on CVD only also demonstrated not to be cost-effective [
41]. Although many economic evaluations have been performed in the field of CMD, none of these studies assessed a prevention program for the combination of these diseases [
10]. Therefore, it remains difficult to directly compare our results with international equivalents.
The scenario analyses showed that it is very hard to reach cost-effectiveness with the evaluated program. Reducing overall intervention costs with 95% (as shown in Additional file
4) to around €7 compared to about €132 as shown in the INTEGRATE trial is not a realistic goal. Non-response is a designated pitfall of stepwise screening programs, as lack of compliance in different steps might reduce cost-effectiveness. However, optimizing response rates in our study would not have resulted in a cost-effective program, due to relatively high intervention costs.
The most promising way to optimize cost-effectiveness is to introduce more effective lifestyle interventions, especially focusing on smoking cessation. The results of our trial showed no significant effect on smoking [
12]. In absolute numbers, 31 of 161 smokers quitted in the intervention group versus 21 out of 161 in the control group. To achieve cost-effectiveness (ICER €20,000 per QALY gained), an additional 18 smokers in the intervention group should have quitted, requiring about three times the effect that was achieved with the current program [
12].
In recent years, there has been more attention for the prevention and treatment of CMD in clinical practice due to renewed guidelines and chronic disease management programs. Ongoing individual case finding might lead to a lower prevalence of undetected high-risk individuals for CMD over time [
42]. Because individual case finding- and structured early detection strategies are fishing in the same waters, this trend may dilute the detection rate of the program and subsequently reduces its cost-effectiveness.
Implications for clinical practice and further research
Stepwise CMD prevention in primary care followed by subsequent treatment appeared not cost-effective. Future research should focus on effective lifestyle interventions and the long-term maintenance of its health benefits, especially focusing on smoking cessation interventions.
It is generally assumed that the higher the risk in patients, the more favourable the cost-effectiveness of preventive procedures. Therefore, alternative strategies to identify high-risk individuals might be promising. Dalsgaard and colleagues found that opportunistic screening for DM2 during a regular GP consultation stimulated higher attendance rates. In addition, it was argued that people attending GP practices might have a more unfavourable CMD risk profile and therefore are likely to be at higher risk [
42]. Ideally, this would lead to the identification of more cases at lower costs. However, this hypothesis should be confirmed by future research. Another strategy could be the selection of high-risk individuals based on routine EHR data, improving individual case finding in daily practice.
Alongside targeted and individual case finding approaches, expanding efforts for universal prevention of CMD plays an important role to reduce overall CMD risk in the total population.
Given our results and the fact that CMD prevention programs are already implemented in several countries, it is important that these programs are evaluated to assess whether these are a cost-effective use of resources compared to other interventions to reduce CMD risk.