Cost-utility analysis in chronic kidney disease patients undergoing kidney transplant; what pays? A systematic review

A systematic review was undertaken to identify all published studies relevant to cost utility analysi in CKD patients undergoing kidney transplant. The protocol has been published in Center for Open Science (OSF) (DOI : https://​doi.​org/​10.​17605/​osf.​io/​xhywn) [15]. The review was conducted from October 2018 to March 2019 and initially all publications in English language published up to March 2019 were included in the review. However, later the search was updated to all published articles until March 2020.

Searches accessed the Medline, the Cumulative Index to Nursing and Allied Health Literature (CINAHL), EMBASE, PsycINFO and National Health Service Economic Evaluation Database (NHS EED) databases by using relevant key words (Additional file 1: Text box 1). The search included only the published journal articles. Previously published systematic reviews on CUA were used to identify the search terms [5, 16].

Further, the reference lists of retrieved articles and review articles in this field of research were searched to identify additional published articles that met predefined inclusion and exclusion criteria (see Fig. 1). The review only focused on interventions in patients undergoing kidney transplant, thus studies of post kidney transplant patients and studies only compared different kidney replacement therapies were not included in the review.

Data extraction was conducted by two independent reviewers (SS and SK) and discrepancies were resolved by discussion. Data fields extracted included research question, study population, setting and location, study perspective, intervention and the comparator, time horizon, discount rate, structure of the economic model, model assumptions, incremental cost and utility, sensitivity analysis—method and the results, characterizing heterogeneity, value of information analysis, Budget Impact Assessment and conclusions.

All monetary values were adjusted to 2016 USD by using CCEMG—EPPI-Centre Cost Converter (https://​eppi.​ioe.​ac.​uk/​costconversion/​). When this information was not reported, it was assumed to be 1 year before publication. The same US dollar value reported in the study was taken if the study was done after 2016. This adjustment is a two-stage process. In the first stage, using a Gross Domestic Product deflator index (‘GDPD values’), original estimate of cost from the original price year is converted to the target price year. In the second stage, the time adjusted cost estimate is converted from the original currency to a target currency, using conversion rates based on Purchasing Power Parities for GDP (‘PPP values’). Thus, this two stage process accounts for both inflation with in the country and fluctuation in exchange rates.

The CHEERS checklist was used as a set of good practice criteria for decision analytic modelling in determining the reporting quality of the economic evaluation. CHEERS check list has been increasingly used to assess the reporting quality of economic evaluations [17‐19]. It has 24 criteria; assessing the title (01 criteria), abstract (01 criteria), background and objectives (01 criteria), methods (14 criteria), results (04 criteria), discussion (01 criteria) and other (02 criteria) (see Additional file 1: Table S1). Each item in the CHEERS checklist was scored as having met the criteria in full (“1”), not at all (“0”) or not applicable (NA). When items partially met the criteria, they were scored as “0”: no partial scores were assigned to avoid introducing subjectivity.

The quality of the data used to inform model parameters was determined using the modified hierarchies of data sources for economic analyses [20]. Each component of the decision model was assessed: baseline clinical data, costs and utilities. The quality of data sources was ranked from 1 to 4 (1 to 5 in cost data) with the highest quality of evidence ranked 1 (see Additional file 1: Table S2). For baseline clinical and cost data used in the evaluations, data from case series or analysis of reliable administrative data bases from the same jurisdiction were considered best quality evidence (rank 1), whereas expert opinion was considered the lowest quality evidence (rank 5 in baseline clinical data and rank 4 in costs data). For utility data used in the evaluations, either direct utility assessment for the specific study from a sample or indirect utility assessment for specific study from patient sample with the disease of interest were considered best quality evidence (rank 1) whereas utility valued derived from Delphi panels or expert opinion were considered the lowest quality (rank 4). In each of the component (i.e. baseline clinical data, costs and utilities), articles ranked either 1 or 2 were labeled as “High quality”, rank 3 as “Medium quality” and rank 4 or 5 as “Low quality”. This classification was carried out in accordance with the previously published literature [3]. For each article, the highest level of evidence used for each parameter was recorded.

Both assessment of the quality of the economic evaluations and assessment of the quality of the data used in each evaluation were conducted by two independent reviewers (SS and SK) and discrepancies were resolved by discussion.

Table 1 shows how each of the 3 categories of articles in the review mapped to each criterion in the CHEERS checklist (see Additional file 1: Table S1). Out of the 24 characteristics, more than 80% of the selected studies reported 14 of the characteristics. The title clearly described the study as an economic evaluation only in 50% (n = 08) of the studies while a structured abstract with all the necessary information was reported in only 56.3% (n = 09) of the studies. The ‘choice of model’, which describes the specific type of decision analytic method used, ideally with a figure to show the model structure, and ‘model assumptions’ which describes all structural or other assumptions underpinning the decision-analytic model, were reported in 75.0% (n = 12) and 62.5% (n = 10) of the studies respectively.

Table 2 describes the quality of the data used to inform model parameters was determined using the modified hierarchies of data sources for economic analyse (see Additional file 1: Table S2). The clinical, cost and utility data used in the three studies in the ‘transplanting infectious kidneys’ category were generally high quality (rank 1 or 2). The clinical and cost data had been gathered from reliable administrative databases or from published evidence from the same jurisdiction where the evaluation was conducted. Both clinical and utility parameters were of low quality (rank 4) in one of the studies in the above mentioned study category [27]. In both ‘CUA of kidney allocation policies, and ‘CUA of technology used in KT’, all the data sources used were generally considered to be of high quality (rank 1 or 2).

Almost all (15/16) [21, 22, 24‐36] the studies included in the review were from high income countries [38], while only one was reported from lower-middle income countries (Malaysia) [23, 39]. Of the 15 reviews reported in the developed countries, seven studies were from United States (USA) and four were from Canada. Of the 16 studies included, three were related to CUA of transplanting infectious kidneys [26, 27, 34], nine were related to CUA of kidney allocation policies [21‐24, 29, 31, 33, 35, 36] and four were related to CUA of technology used in KT [25, 28, 30, 32].

Table 3 provides an overview of the study characteristics of the reviewed models. Twelve studies were performed from a “healthcare payer perspective”, while five were from the societal perspective. The discount rate used ranged from 1.5 to 5% while the time horizon varied from 5 years to lifetime. Markov decision modelling was used in majority of the evaluations (13/16). Though the model structure used had not been explicitly mentioned, evidence of use of a Markov model was identified in two studies [23, 27].

Table 4 provides the results of the CUA included in the review. The authors of 13 studies reported the intervention as cost saving and improving quality of life [21‐28, 30, 32, 34‐36], whereas three studies were cost increasing and improving quality of life [21, 24, 27]. Transplanting a Hepatitis C nucleic acid test positive deceased donor kidney followed by post-transplant direct acting anti-viral administration to the recipient, screening of all donors for HCV status and transplanting infected organs into HCV⁺ recipients, ignoring the HCV status of the donor when transplanting and transplanting kidneys from donors who are at increased risk of developing HIV or Hepatitis C were found to be cost-saving as well as delivering increased QALY. Further, several kidney allocation interventions were found to be cost-effective. Transplanting HLA 0‐3 or 4–6 mismatch live donor kidneys, a policy of transplanting the top 20% of the Kidney Donor Profile Index kidneys to candidates in the top 20% of expected survival, live donor transplantation, including both donation after brain death and donation after cardiac death kidneys in the allocation pool, the Programme Alba [40] implemented in Italy, a payment of US $8,000 (2010) to all the living donors which would expect the annual transplant rate to increase by 5%, were also found to be cost-effective. Using bead-based multiplex assays (threshold Mean Fluorescence Intensity level 500) with Complement-Dependent Cytotoxicity in screening for donor-specific anti-HLA antibodies to determine transplant suitability, use of flow screening only where patients’ immunological risks were stratified using the results of Flow Cytometry Cross Matching and flow micro-bead Panel Reactive Antibody when assessing the HLA antigen status, using hypothermic machine preservation as the organ preservation method in KT and pre-operative imaging of live kidney donors using Digital Subtraction Angiography were cost effective interventions. The intervention, HLA-B locus not matching before kidney allocation, was found to be cost saving while losing QALY [31]. The willingness to pay threshold (WTP) was reported in only two [21, 26] of the studies.

In addition to the baseline analysis, sensitivity analysis was performed in all the evaluations except one [21]. Sensitivity analysis provides the information on the robustness of the baseline results according to different parameter estimates or, putting it another way, characterizes the effect of uncertainty in model parameters on the results [3, 41]. Of the evaluations which performed sensitivity analysis, 12 have performed deterministic sensitivity analysis [42], either one-way or two-way sensitivity analysis. Four studies [26, 27, 31, 34] have reported results of scenario analysis while one study [25] reported results of bootstrapping. Five evaluations that characterized parameters as distributions used probabilistic sensitivity analysis [22, 26, 31, 32, 36]. The results of the sensitivity analysis were robust in 15 of the evaluations, while the results were variable in five. In one-way, two-way or scenario analysis, if the baseline results do not significantly change or in probabilistic sensitivity analysis (PSA) if more than 50% of the iterations confirm the baseline results, the results of the sensitivity analysis was considered robust. Kadatz et al. [26] had gone one-step further and included information on Budget impact Assessment. However, none of the articles reported on Value of Information analysis.

To assess the reporting quality of the economic evaluations, the 24 criteria in the CHEERS checklist were used as the benchmark good practice criteria for decision analytic modelling. According to the CHEERS guideline the title should reflect that the study is an economic evaluation and the abstract should provide a structured summary of the evaluation. The current review identified that the title and the abstract of most of the reviews selected was poor in quality. Similar findings have been found by Rosen et al. where it was found that the abstracts published in economic evaluations frequently omit information critical to proper study interpretation [43]. It is imperative that the title and the abstract contain all the necessary information because this is the only information accessible in some settings, especially in lower and middle income countries. In these jurisdictions health policy makers, planners and clinicians are forced to make decisions based on the subset of information in the abstract [44].

Perspective of the study is the viewpoint from which the intervention’s benefits and costs are evaluated. It is said that the societal perspective for economic evaluation is the ideal approach in assessing the profitability of societal investments (eg: kidney transplant) [45]. Eleven of the 16 studies included in the review used only healthcare payer’s perspective, five studies used societal perspective while only one study assessed both healthcare payers’ and societal perspective. Most of the interventions evaluated in the review not only incur cost to the healthcare payer but, in the border context, to the society as well (eg: productivity loss, premature deaths). Thus, narrow economic perspectives used in most of the evaluations limits the usefulness of these evaluations by excluding relevant costs and health outcomes from the analysis. Arguably for a chronic disease like CKD and its treatment the time horizon used in the model should be long enough to capture the chronic sequelae of the disease. Except Cavallo et al. [24], all the other studies had used 10 years or more as the time horizon and three studies have applied a life-time horizon for the CUA. However, these three studies have not explicitly stated the number of years that they have considered as “life-time horizon”. Considering the fact that mortality rates among post kidney transplant patients, even after a successful kidney transplant, is higher than the general population, explicitly stating the time horizon could be more informative.

Assumptions used in an economic model have a significant impact on model prediction [46]. The model assumptions were clearly reported in only 10 of the studies included in the review, which could adversely affect the transparency of some of the models used. Further, the level of confidence in the results of the economic evaluation can be boosted by sensitivity analyses. Of the several methods of sensitivity analyses, PSA provides the strongest evidence of robustness of the results, as it explicitly indicates the probability that an intervention is cost-effective [46]. When evaluating the results of this review, it is important to bear in mind that PSA has been performed only in five of the papers.

Stakeholders’ willingness to pay (WTP) threshold was reported only in two of the evaluations. WTP is an important characteristic that helps payers deciding how to weight resource allocation, in particular when interventions are more effective but more costly [47]. Value of information analysis, which estimates the added value of future cost-effectiveness research [48], Budget Impact Assessment, which assesses the financial consequences of adopting an intervention within a specific context and economic evaluations in decision making [49] are sophisticated methods that are gaining wider uptake. Value of information analysis was not reported in any of the evaluations reviewed while Budget Impact Assessment was reported in only one [26]. This limits the comprehensive understanding of the economic assessment of the new interventions discussed in this review.

The quality of data used in the models is an important factor when making a decision about the cost-effectiveness of one intervention over another. The quality of data used in the models is high. Most of the clinical and cost data were extracted from reliable data bases, which make the data sources high in quality. Almost all the studies used published data for the utility scores.

Though this study identified a substantial number of interventions for the KT population that appeared to be cost effective, it requires a robust systematic implementation to translate some of the interventions into practice. Currently safe and highly effective direct-acting antiviral (DAA) therapy is available for hepatitis C virus and recent evidence has demonstrated excellent safety and efficacy of DAA therapy in renal transplant recipients. Thus it is anticipated that recipients of HCV infected kidneys will have improved long-term graft and patient survival in the future [50‐52]. Having said that, between 2005 and 2015 around 4000 donor kidneys have been discarded in United States for being positive for HCV, and it is expected that the numbers will fall after liberal use of DAA in the future [53]. Further, currently the recipient’s consent is needed to transplant a hepatitis C-positive organ and recipients’ willingness to receive infected kidneys has not been systematically evaluated. It is encouraging to note that the recent amendments to the United States’ law now permits people living with HIV to donate organs to HIV-infected recipients under research protocols [54]. This legislative effort has the potential to expand the donor pool by about 100 kidneys per year in the United States in the future [55].

Two of the studies in the reviews indicated that a payment to the living donors is a cost-effective strategy to increase the donor pool. Payment for organs is perhaps one of the most controversial interventions suggested to increase the kidney donor pool. It has generated disagreement on legal grounds and concerns about public acceptance [56]. At the core of the disagreements are moral and ethical concerns of commodification of the body and the risk for coercion leading to wealth-based distortions in decision-making [57‐59].

Several kidney allocation policies (eg: a policy of transplanting the best quality kidneys to candidates with the highest expected survival, patients receiving a paid living unrelated donor kidney) were found to be cost effective in the review. But these allocations disadvantage certain groups eg elderly transplant candidates, fueling a debate about the equity of the new policies [60, 61]. Healthcare policy makers have particularly difficult decisions in this area, reconciling gain in quality of life associated with a specific change to allocation policy against any loss of equity to certain patient groups [56].

Though broad search strategies were used to find the relevant articles, we may have not identified all the model based economic evaluations reported in the specific area. Further, we found 20 abstracts presented in conference proceedings, in which the full article was not available for the review and were therefore not included. Our assessment of the reporting quality of the economic model and the quality of the data used in the model were based on the way the evaluations were reported rather than conducted. Thus, the conclusions should be interpreted with caution. Finally, an intrinsic limitation connected to assessment of quality (reporting quality and data quality) is that the results of the quality assessment may vary depending on the assessor. To minimize this potential bias, all studies were evaluated by the same two researchers and discrepancies were resolved by discussion.