Costs of clinical events in type 2 diabetes mellitus patients in the Netherlands: A systematic review

Background Type 2 diabetes mellitus (T2DM) is an established risk factor for cardiovascular and nephropathic events. In the Netherlands, prevalence of T2DM is expected to be as high as 8% by 2025. This will result in significant clinical and economic impact, highlighting the need for well-informed reimbursement decisions for new treatments. However, availability and consistent use of costing methodologies is limited. Objective We aimed to systematically review recent costing data for T2DM-related cardiovascular and nephropathic events in the Netherlands. Methods A systematic literature review in PubMed and Embase was conducted to identify available Dutch cost data for T2DM-related events, published in the last decade. Information extracted included costs, source, study population, and costing perspective. Finally, papers were evaluated using the Consolidated Health Economic Evaluation Reporting Standards (CHEERS). Results Out of initially 570 papers, 36 agreed with the inclusion criteria. From these studies, 150 cost estimates for T2DM-related clinical events were identified. In total, 29 cost estimates were reported for myocardial infarction (range: €196-€27,038), 61 for stroke (€495-€54,678), fifteen for heart failure (€325-€16,561), 24 for renal failure (€2,438-€91,503), and seventeen for revascularisation (€3,000-€37,071). Only four estimates for transient ischaemic attack were available, ranging from €587 to €2,470. Adherence to CHEERS was generally high. Conclusions The most expensive clinical events were related to renal failure, while TIA was the least expensive event. Generally, there was substantial variation in reported cost estimates for T2DM-related events. Costing of clinical events should be improved and preferably standardised, as accurate and consistent results in economic models are desired.


Objective
We aimed to systematically review recent costing data for T2DM-related cardiovascular and nephropathic events in the Netherlands.

Methods
A systematic literature review in PubMed and Embase was conducted to identify available Dutch cost data for T2DM-related events, published in the last decade. Information extracted included costs, source, study population, and costing perspective. Finally, papers were evaluated using the Consolidated Health Economic Evaluation Reporting Standards (CHEERS). PLOS  Introduction Type 2 diabetes mellitus (T2DM) is an established risk factor for vascular complications, cardiovascular events and renal failure [1,2]. Also, T2DM is the most prevalent chronic disease in the Netherlands. In 2014, an estimated 960,000 patients with T2DM were known to the general practitioner, which is about 5.7% of the Dutch population [3,4]. The prevalence is expected to be as high as 8% in the year 2025 [5]. Besides significant clinical impact, this will result in profound increases in healthcare expenditures and highlights the need for appropriate assessment of T2DM drugs' cost-effectiveness and well-informed reimbursement decisions. T2DM treatment is aiming to normalise blood sugar levels, blood pressure and lipids with the ultimate goal to prevent cardiovascular and renal complications. Major cardiovascular complications include myocardial infarction (MI), stroke, transient ischaemic attack (TIA), heart failure (HF), and revascularisation. Major renal complications include end-stage renal disease (ESRD), dialysis, and kidney transplantation. Given their significant impact on both patients' health status as well as healthcare expenditures, a prerequisite for state-of-the-art health-economic evaluations is the full understanding and consistent use of T2DM-related complications' costs, especially those of related cardiovascular and nephropathic events.
In the Netherlands, the National Health Care Institute (Zorginstituut Nederland, ZIN) provides guidelines for pharmacoeconomic research, highlighting key methodological issues that should be addressed for an adequate economic evaluation [6]. For instance, according to the guideline, economic evaluations should be carried out using a societal perspective, taking into account costs both inside and outside the healthcare system, and thus including e.g. productivity losses. Indeed, according to pharmacoeconomic guidelines, costs for clinical events are essential for designing adequate and valid health-economic models [6,7]. Yet, availability, a standardised measurement, and consistent use of costs of clinical events related to T2DM is limited. Here, we aim to systematically review available recent costing data for T2DM-related major cardiovascular and nephropathic events in the Netherlands.

Study design
A systematic literature review was conducted to identify all available publications specifying Dutch costs for clinical events commonly found in T2DM patients. This review was reported according to the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) literature review methodology [8], provided in S1 Table. Search strategy The PubMed database was searched for publications between January 1 st 2005 and January 1 st 2018. An overview of keywords can be found in Table 1. In addition, reference lists of identified reviews and meta-analyses on the topic were searched for potentially relevant articles.
Subsequently, Embase was also searched with adapted keywords from Table 1, corresponding with Emtree terms. In this analysis, results also found in MEDLINE were filterred out, as to improve efficiency.

In-and exclusion criteria
Articles were included in this review if they met the following criteria: The papers considered direct costs for the specified clinical events per patient in the Netherlands.
Papers required to be published between January 1 st 2005 and January 1 st 2018, to reflect recent data only, as older estimates may be "outdated" and irrelevant for present day.
The paper evaluated at least one of the six predefined major clinical events (MI, stroke, TIA, HF, renal failure, and revascularisation). Papers included T2DM patients or patients with clinical events commonly associated with T2DM. It should be noted that these clinical events are not restricted to T2DM patients and can occur in patients without T2DM as well. Therefore, given the focus on the costs for these clinical events, the patients in the included studies were not always T2DM patients.
The full-text of the paper required to be accessible. Articles reflecting guidelines or study protocols, meeting reports, or case reports were excluded.

Extracted information
Direct medical costs, such as those for hospitalisation, medication and rehabilitation, were included. Additionally, and in line with the Dutch preferred societal perspective, indirect costs such as productivity losses were identified and included. If comparative studies were identified, the costing in the standard-of-care arm was chosen to be included, as it reflects standard of practice more accurately. Next to costs, data extracted included the cost source, study population, and the costing perspective (e.g. a healthcare payer, the hospital or the society).

Domain Search terms
Subject "costs and cost analysis" OR "cost-effectiveness" OR "cost-utility" OR "cost-benefit" OR "cost-effective" OR "economic evaluation" OR "economic analysis" Events "diabetes mellitus" OR "stroke" OR "myocardial infarction" OR "heart failure" OR "ischemic attack, transient" OR "myocardial revascularisation" OR "albuminuria" OR "acute renal injury" OR "renal insufficiency" To ease comparison, we also reported maximum and minimum annual costs in 2018 euros, with standard Dutch inflation rates used for standardising costs from previous years [9].
For papers that were health economic evaluations, adherence to the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) was also assessed to put the individual cost estimates into perspective, regarding time horizon, reporting perspective, et cetera [10].

Search results
The results of the systematic literature search are displayed in the flowchart in

Overview
In total, 36 studies met our inclusion criteria. From these studies, 150 cost estimates for clinical events related to T2DM were identified. All papers provided the year of costing, except for three [11][12][13]. For the first two studies, the year of costing was assumed to be the year before their respective publication, while for the latter, the year of costing was found in one of its references. Table 2 provides the characteristics of the included studies. Nineteen analyses took a healthcare payer perspective, seven a societal perspective, six a hospital perspective, one a third-party payer perspective, and three articles did not specify the perspective used. Of all studies, eight explicitly stated that in addition to direct costs, indirect costs were also taken into account [1,12,[14][15][16][17][18][19]. Four of these studies took productivity losses into account, and these four all used a societal perspective [1,14,16,18]. The other four studies took overhead costs into account, although these are not considered indirect costs in health economics.
Five studies used the Diagnosis Treatment Combination (Diagnose Behandel Combinatie, DBC), the Dutch case-mix categorisation aligned with resource use and applied for reimbursement of hospitals [20][21][22][23][24]. These DBCs are comparable with diagnosis related group (DRG) based systems used in other countries, although certain differences do exist, such as goal and scope [25]. In three studies, information on resource use was gathered from the EDISSE trial [13,26,27], while the sources for the other papers concerned registries, case record files, cost diaries, trials, billing systems, or cost estimates from older costing studies.
The Dutch Manual for Costing in Economic Evaluations was referenced by eighteen studies (62%), eleven of which used the manual to derive standard prices. Tariffs provided by the Dutch Healthcare Authority (Nederlandse Zorgautoriteit, NZa) were used in four studies [20,23,28,29]. Furthermore, nine articles gathered unit costs from hospitals directly. Fourteen studies used at least one costing study to derive their cost estimates from. Of these fourteen studies, it was found that nine referenced at least one paper published before 2005, and four references reported their costs in Dutch guilders. The cost estimates derived from one paper were reported in 2002 US dollars [30]. These estimates were converted to 2002 euros, using an exchange rate of 1 EUR = 0.95 USD [31].

Costs for clinical events related to T2DM
Considerable variation among the reported costs was found.

Myocardial infarction
In seventeen studies, 29 different cost estimates for MI were used, with costs calculated between 1999 and 2015 ( Table 3). The papers used various methods to derive cost estimates, such as hospital database analyses, and expert opinions. It was possible to make a differentiation between first-year (acute) and follow-up costs. Seven studies also considered the follow-  up costs after year one [1,24,28,29,36,37,43]. For patients with MI, the average costs in the first year ranged from €3,390 to €27,038 per patient. The costs accrued in the subsequent years ranged between €1,026 and €2,092 per patient per year. The studies that defined which direct costs were covered only took hospitalisation costs into account, ergo no rehabilitation or other costs were covered. Three papers also included indirect costs, though only one accounted for productivity losses [1], as the other two included only overheads [15,17].

Stroke
The majority of papers reported cost estimates for stroke events: 61 cost estimates in 22 studies were identified ( Table 4). The year of costing ranged from 1999 to 2015. Four papers reported on the different severities of stroke [16,28,38,42], and four papers mentioned the type of stroke studied: four reported on ischaemic strokes specifically [16,20,38,47], while one reported haemorrhagic strokes in addition [38]. Four studies reported cost estimates for the first year separated over two periods of six months [18,26,37,42]. One paper reported cost estimates per 3-month cycles [38], and another for just the first six months [13]. Two studies showed that the majority of the cost estimates for stroke events were made up by hospital stay [15,20]. When rehabilitation was considered, it made up an even larger share than hospital stay [18,21]. Three studies also included indirect costs [1,15,18]. Although Tholen et al. did take indirect costs into account, the cost estimate in Table 4 does not include productivity losses, since the study reported them separately [16].

Transient ischaemic attack
Four different costs for TIA were reported in three different studies, with the lowest cost estimate being €587, and the highest amounting to €2,470, reflecting inpatient and outpatient costs, respectively [20]. Hospital stay accounted for the largest share of the costs; €1,748 of the €2,470 were for inpatient stay [20].

Heart failure
A total of fifteen cost estimates for heart failure were reported in eight studies ( Table 6). The costs for heart failure varied between €945 and €16,561 per patient per year. One study reported a cost estimate for fatal congestive heart failure (HF), which was assumed to be 50% of non-fatal congestive HF events [1]. Three papers reported estimates for the subsequent years of congestive HF [1,24,29]. These ranged between €325 and €6,672. One study reported separate costs for different severities of HF, whether HF went undetected, and by sex [45]. The main cost driver for heart failure is hospital stay, both for the first year as well as the follow-up years, where rehospitalisation accounted for 73% of the costs [44].

Renal failure
A total of 24 cost estimates associated with renal failure were reported in seven studies (Table 7). Several different types of dialysis were covered in five papers [14,24,29,32,34], reporting fifteen cost estimates in total, which ranged between €54,067 and €89,447. The cost estimates for dialysis in subsequent years were reported to be equal to the estimates for the first year. Three studies reported five cost estimates on renal transplantation [24, 29,32]. First year cost estimates ranged between €14,387 and €91,503, and costs for subsequent years were estimated at €2,438 and €3,680. Four cost estimates for end-stage renal disease (ESRD) were reported in three studies [1,32,41]. These costs ranged between €3,640 and €69,440. Adarkwah et al. calculated a weighed mean for ESRD patients, taking renal transplantation and different  types of dialysis into account [32]. The costs for one year of ESRD were estimated at €42,219, while one year of dialysis and renal transplantation cost €79,112 and €14,387, respectively.

Revascularisation
A total of seventeen cost estimates for revascularisation were reported in nine studies (Table 8). Two types of revascularisation were assessed; percutaneous coronary intervention (PCI) and coronary arterial bypass grafting (CABG), with respectively ten and seven cost estimates identified. PCI cost estimates ranged from €3,000 to €14,037 [37,40]. For CABG, the cost estimates ranged between €5,441 and €18,010 [33,40]. Table 9 presents the results of the papers that were assessed according to the CHEERS guideline. Overall, papers' adherence to the checklist was found to be high, even though articles did Costs of clinical events in T2DM patients in the Netherlands: A systematic review not explicitly state whether a reporting guideline was used. The items with the lowest amount of adherence were measurement and valuation of preference-based outcomes, assumptions, and characterising heterogeneity. Assumptions were not clearly defined in five studies [13,14,21,35,39], and only partially in two [44,47]. Finally, one study failed to characterise heterogeneity [39].

Main findings
In this systematic review, we aimed to outline the Dutch cost estimates of six-major T2DMrelated clinical events. It was found that many studies reported on cost estimates for MI and stroke, but only a limited number focussed on other T2DM-related clinical events. The most expensive clinical events were found to be related to renal failure, most notably ESRD and dialysis, although some estimates reported stroke and CABG to be a significant source of expenditures as well. MI, TIA, and HF were generally among the least expensive T2DM complications. A large variety in cost estimates was found in the included studies. Adherence to CHEERS guidelines was generally high.

Interpretation
While some heterogeneity is to be expected when dealing with estimates, most values showed poor agreement, sometimes even between a study and its reference. For example, Verhoef et al. reported cost estimates for TIA, but this value was considerably lower than the cost estimates found in their sources [47,61]. Three studies reported cost estimates for stroke separated into minor and major stroke [16,28,38]. In 2018 EUR-corrected values, minor stroke was estimated to cost €9,079, €7,342, and €24,557, respectively, during the first year of stroke and €1,554, €1,256, and €6,177 in Table 6. Summary of studies reporting costs for heart failure.

Study
Specific type Initial Follow-up Costs covered As is evident from these values, cost estimates for minor stroke, both in the first year and subsequent years, differed substantially. In contrast, cost estimates for major stroke showed more agreement. A possible reason for these discrepancies could be the definitions of minor and major stroke between the studies. Jacobs  While fifteen studies did manage to evaluate resource use and resource costs by means of questionnaires, record files and databases, eleven studies in this review cited sources predating 2005, some even reporting cost estimates in Dutch guilders instead of euros. Estimates derived from older papers may give rise to costs not representative of current costs found in healthcare. Furthermore, this may give rise to a risk of bias, as previous research could be unable to meet the specifications needed, whereas costs derived from first-hand sources, e.g. hospital records, are seen as a more accurate reflection.

Unit cost year 1 in € (year) Source(s) Unit cost year 2+ in € (year) Source(s) Direct costs Indirect costs
Although the Dutch guidelines for economic evaluations prefer the societal perspective [6], only five studies actually took this approach. Moreover, one of these papers explicitly stated that even though the societal perspective was used, the indirect costs due to lost productivity losses were not accounted for, because of the advanced age of the patient group [44]. This means that even though the cost estimate was derived using a societal perspective, it is in fact an incomplete value, considering travel costs for patients and caregivers are accrued regardless of patient age. Regarding the other papers, either a hospital perspective or a healthcare perspective was used. These perspectives lack direct non-medical costs, such as travel costs, as well as indirect non-healthcare costs. Therefore, these cost estimates lack societal costs such as productivity losses.
Tan et al. was the only paper included in this review that compared different costing methodologies [15]. In their paper, bottom-up microcosting, top-down microcosting, and gross costing were compared, with gross costing differing the most. If other costing studies had specified their costing methods, it could provide insight into the reason for the heterogeneity found in cost estimates.
In a systematic review about the costs of treating cardiovascular events in Germany, Schmid stated that 80-85% of costs in the first year after MI are spent in the first six months, meaning the other six months of the first year make up for just 15-20% [74]. For the Netherlands, one paper reported first year cost estimates separated in two periods of six months [37]. They found that the cost estimates for the first six months were €10,250 and €2,500 for the subsequent six months. This means that around 80% of the costs in the first year after MI are incurred during the first six months in the Netherlands, similar to Germany. Comparable findings were found for stroke events. In the same paper, Schmid reported that in Germany, 80% of the costs for stroke during the first year were reached within the first half-year. Three papers reported cost estimates for stroke in the Netherlands in periods of six months [18,26,37]. These studies show that between 69 and 74% of the costs during the first year are incurred in the first half-year. From these results, it is clear that the larger part of the first-year costs associated with MI and stroke are incurred within six months of the event. This is to be expected, considering hospitalisation and rehabilitation are the main cost drivers, and are mainly present immediately after an event.
The fact that only three studies reported on the costs associated with TIA illustrates the systematic underreporting of TIA. It being a transient event, a TIA can go unnoticed, even though the patient is at a higher risk of other cardiovascular events, or it is grouped with strokes because of its definition, resulting in an increased number stroke reports, but also decreasing the average reported costs for stroke [75].
A possible cause for the difference in costs found between studies could be the developments in healthcare. For example, more efficient or less expensive procedures could have become the standard. For us, this means that adjusting for inflation would not be enough to carefully compare costs from 2005 to costs from 2015.
To provide context of the studies in which the costs were used, we also reviewed each study using the CHEERS checklist. While the majority of CHEERS items focus on more methodological issues, some items of CHEERS were particularly relevant for this study. These items were "Estimating resources and costs" and "Currency, price date, and conversion". Generally, the first item was well-addressed, while in the latter, some room for improvement was noted, as either the price date or the conversion method was not mentioned [11,13,16]. However, as the CHEERS statement was developed as a guideline for the reporting of health economic evaluations, the quality of cost estimation cannot be adequately assessed solely with this checklist.

Strengths and limitations
To the best of our knowledge, this is the first systematic review specifically focussing on Dutch costs of six major T2DM complications. Results can be of relevance for future cost-effectiveness analyses of new type 2 diabetes treatments in the Netherlands. However, also certain limitations have to be acknowledged. Due to our strict inclusion criteria and focus on major cardiovascular and renal complications, no attention could be paid to other T2DM-related events, such as unstable angina, peripheral artery disease, neuropathy, diabetic foot, and retinopathy. Furthermore, the costs associated with micro-and macroalbuminuria were not explicitly reported, although these cannot be considered as events, but rather as bio-factors or risk factors for renal events, and if relevant, these were included in the costs for ESRD. Other T2DM complications are recommended to be included in more comprehensive future studies.
Finally, as most cost estimates reported in the identified studies were based on models or costs derived from guidelines, instead of trial-based values, the generalisability is limited. This means that variance and representativeness of patient samples that make up cost estimates could not be evaluated. Therefore, our review focussed on a descriptive analysis of our findings.

Conclusions
This systematic review showed that there is substantial variation in reported cost estimates for six major complications associated with T2DM. Most of the studies reported on MI and stroke. Due to a limited amount of papers covering heart failure, revascularization, TIA and renal failure, cost estimates varied widely and transparency regarding cost sources was generally poor. The costing of clinical events related to T2DM should be improved and preferably standardised, if accurate and consistent results in economic models are desired.
Supporting information S1 Table.