Literature Search Strategy

A literature search was conducted via the OvidSp interface on January 22, 2014 using keywords for HAT specific to the Medical Subject Headings (MeSH) terms required for Medical Literature Analysis and Retrieval System Online (MEDLINE) and Embase databases. An economic filter developed by Scottish Intercollegiate Guidelines Network (SIGN) was also applied. (Refer to S1 Supporting Information) The Journal Storage (JSTOR) database was also searched using the following key words: African trypanosomiasis OR trypanosom& OR “sleeping sickness” AND cost& AND economics. In addition, the following keywords were also searched in the Database of Abstracts of Reviews of Effects (DARE), National Health Service Economic Evaluation Database Health Technology Assessment (NHSEED HTA), and Cochrane databases: “African” AND “Trypanosomiasis” OR “sleeping sickness”. All citations were downloaded into Mendeley, where duplicates were identified and removed.

Literature Screening & Inclusion/Exclusion Criteria

Screening of the articles was done in two stages. At the first level, all titles and abstracts were screened. Articles that were considered potentially relevant were then assessed at the second level, in which the full text was read. After reading the full text, articles that still met the inclusion criteria were considered. A full description of the inclusion and exclusion criteria is available in S2 Supporting Information. Data were screened on both levels according to the outline of population-intervention-comparators-outcomes-setting (PICOS) criteria, in which the population pertained to humans. Evaluations regarding strains of both HAT T. b. gambiense and T. b. rhodesiense were reviewed, although outcomes only pertaining to humans impacted by the disease were taken into consideration (no animal implications). Interventions (I) and comparators (C) included any intervention that could lead to prevention or reduction of disease in human populations (including vector control). The outcomes (O) that were considered for review were costs, consequences (life-years saved [LYS], disability-adjusted life years [DALYs], etc.), and the incremental cost-effectiveness ratio (ICER), while the setting (S) included any African country. For the purpose of this analysis, an economic evaluation was defined by the Drummond et al. definition of a “full economic evaluation,” and therefore, both costs and consequences of two or more alternatives had to be present in the analyses evaluated [13]. In cases in which an incremental analysis was not performed, articles were not excluded. Instead, if there was sufficient information in the publication to calculate the ICER, it was calculated during the review process. If there was insufficient information to calculate the ICER, it was noted in the critical appraisal that an incremental analysis was not present. No time constraints were added to the search.

Quality Assessment and Critical Appraisal

The quality of the included studies was assessed using the SIGN Methodology Checklist 6: Economic Evaluations Version 3.0 [14], which was composed of two parts. The first portion contained questions regarding the internal and external validity of the publication. Items in the sections were assessed using answers of “Yes,” “No,” or “Can’t say.” The second portion of the checklist addressed the reviewers overall assessment of the study and also provided the reviewer with an area to judge if the article was “unacceptable,” “acceptable,” or of “high quality.” Studies that received a “Yes” on 65% or more of the questions in Section 1 were considered acceptable to the authors.

Literature Search Results

The NHSEED, JSTOR, MEDLINE, and Embase searches yielded a total of seven articles, 1,000 articles, 595 articles, and 673 articles, respectively. An additional eight articles from the grey literature, reference lists, and referrals from subject matter experts were also included. There were a total of 2,283 studies found, and after the removal of duplicates, 2,095 were chosen for primary screening (title and abstracts). A total of 41 publications were then selected for full-text screening. Thirty-four studies were excluded after full-text review, and reasons for exclusion were recorded. (Refer to Table 1.) Seven full texts [15–21] were included for full critical appraisal and data abstraction for analysis. (Refer to Fig. 1.)

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Characteristics of excluded studies at second-level screening.

https://doi.org/10.1371/journal.pntd.0003397.t001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Preferred reporting items for systematic reviews and meta-analyses (PRISMA) diagram.

JSTOR, Journal Storage; MEDLINE, Medical Literature Analysis and Retrieval System Online; NHSEED, National Health Service Economic Evaluation Database. 1000*: Although 1,490 articles were found using JSTOR, only 1,000 articles were accessible due to limitations of the JSTOR database.

https://doi.org/10.1371/journal.pntd.0003397.g001

Quality Assessment and Critical Appraisal

The quality scores for the seven included studies [15–21] displayed in Table 2 (SIGN Methodology Checklist 6: Economic Evaluations) demonstrated that on average 81% (67%–89%) of the items stipulated by the SIGN checklist were addressed. Economic theory suggests that individuals have a time preference in regards to gains, and hence, costs and outcomes in the future are less valuable than those in the present [22]. This concept is referred to as “discounting” and is standard methodology in economic evaluation; however, five out the seven studies in this review did not address it [16–19,21]. Each publication considered the cost and consequence compared to more than one intervention for HAT; however, three of the publications [15,17,18] did not include an incremental analysis to examine the marginal benefit of adopting one intervention compared to the next best option. A single study [19] did not have a clear objective, and Shaw’s study did not justify the study design or clearly describe the cost sources [15]. All but one study [20] completed a sensitivity analysis in addition to the base results. All studies discussed the economic importance of the question and had outcomes that could be relevant for decision makers. Overall, all studies were judged to be “acceptable” for this review.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Critical appraisal (Scottish Intercollegiate Guidelines Network (SIGN) Methodology Checklist 6: Economic Evaluations).

https://doi.org/10.1371/journal.pntd.0003397.t002

Characteristics of Included Economic Evaluations

Each of the seven included publications had varying characteristics, as summarised in Table 3. The first publication of a full economic evaluation for HAT identified was completed in 1989 by Alexandra Shaw [15], with the next publication coming in 1995 [16]. The remaining five publications were published from 2005 to 2008 [17–21]. The evaluations covered four African countries: Democratic Republic of the Congo (DRC), Uganda, Côte d’Ivoire, and Angola. Most (3/7) evaluations (n = 3) came from DRC [18–20], with one study from Côte d’Ivoire [15], one study from Uganda [16], one study from Angola [21], and finally one study that included an analysis from both Uganda and Côte d’Ivoire [17]. Economic evaluations concerning HAT in human populations looked almost exclusively at the disease T. b. gambiense (71%), although in two instances the disease strain was not specified explicitly [15,21]. A total of four economic evaluations [15,18,19,21] were considered cost-effectiveness analyses (CEA) in which the cost for a desired effect or consequence (e.g., lives saved, years of infection avoided, etc.) was measured. Two studies [16,20] included both a CEA and cost utility analysis (CUA) in which the utility was measured in DALYs. One study exclusively completed a CUA in which cost per DALY averted was measured as the main outcome [17]. Overall, there was only one publication that was found in an “economic” journal, as the remaining articles were published in journals pertaining to tropical medicine and infectious diseases. Funding for the research was often not mentioned. However, WHO was referred to as a means of support in two publications [16,18], and support from the Belgian Directorate General for Development Cooperation was also mentioned [18].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Characteristics of included economic evaluations.

https://doi.org/10.1371/journal.pntd.0003397.t003

Interventions

The majority (5/7) of the publications evaluated interventions that included case detection and diagnosis, while two of the articles evaluated treatment interventions of melarsoprol and eflornithine (difluoromethlyornithine [DFMO]) for stage 2, as the treatment for stage 1 was always considered to be pentamidine [16,21]. Two publications by Lutumba [18,19] looked exclusively at sensitivity and specificity of diagnostic algorithms and staging algorithms, while one study also looked at the differences between treatment and vector control interventions in addition to case detection and diagnosis [15]. The study by Shaw in 1989 was the only publication that included a comparative economic analysis for vector control as an intervention to control HAT in a human population.

Economic Evaluation Description

Key insights regarding the details of the included economic evaluations are described below and also summarised in Table 4.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Description of included economic evaluations.

https://doi.org/10.1371/journal.pntd.0003397.t004

Methods and Software

Six of the seven included studies used modelling to measure outcomes for the economic evaluation. Only one study completed an economic evaluation alongside a clinical trial. The most common form of modelling was decision tree modelling; the structure of the remaining models was not described in detail although they were all described as being implemented with spreadsheets. For decision tree models, TreeAge software (TreeAge Software, Williamstown, Massachusetts, United States) was used for three of four studies [18,19,21], and one publication did not mention which software was used. The two spreadsheet models that were reviewed [15,17] used Super-Calc 4 (Sorcim, Silicon Valley, California, US) and Microsoft Excel (Microsoft Corp., Redmond, Washington, US) software, while the economic evaluation alongside clinical trial (EEACT) [20] relied on Microsoft Access (Microsoft Corp., Redmond, Washington, US), Microsoft Excel (Microsoft Corp., Redmond, Washington, US), and Epi Info 2002 (Centers for Disease Control and Prevention, Atlanta, Georgia, US).

Model Structure, Assumptions & Validation

A visual diagram of the model was provided for five of the six studies that included models [16–19,21]. Although descriptions of the six models were available, no details of the assumptions or justification for the inputs used in the modelling were addressed in any of the included literature. None of the articles reported completing an internal validation of the models, but the authors of one article [19] did compare their outcomes to other literature in similar areas for external validity.

Population, Setting, and Perspective

In one of the modelling studies, the number of patients modelled was not mentioned, while the remaining studies included 690 to 1,000,000 hypothetical patients. The clinical trial included a total of 57 patients from 47 households [20]. As mentioned previously, the populations were based on four countries (DRC, Côte D’Ivoire, Angola, and Uganda), with different settings including: rural communities, health centres, and a sleeping-sickness hospital ward.

In one case [15], the perspective of the analysis was not mentioned, but two articles approached the economic evaluation from a societal perspective [16,20], and the remaining four articles used the provider perspective (e.g., a donor or national health service) [17–19,21].

Additional Inputs, Outcomes, and Features of Included Economic Evaluations

Data sources for the economic evaluations came from clinical trials, primary data collection from national programmes (e.g., Programme National de Lutte contre la Trypanosomiase Humaine Africaine [PNTHLA], Médecins Sans Frontières [MSF], and National Sleeping Sickness Programme Uganda), reports from WHO, available literature, and from speaking with experts in the arena of HAT. Prevalence values were not mentioned in two studies and ranged from 0.1% to 70% in the remaining literature.

All costs were evaluated in US dollars (USD} [15–18,20,21] except for one study by Lutumba et al. [19] that estimated cost-effectiveness in euros. Three studies reported only one outcome, while the remaining studies reported two outcomes in terms of cost per outcome. Cost per DALY averted was reported in three studies, while cost per LYS was reported in four studies. Cost per years of life lost (YLL), cost per patient/control case detected or patient cured, and cost per infection prevented were also examples of cost-effectiveness reported in the literature reviewed. Shaw (1989) and Shaw and Catt and reported time horizons of 20 years and one year, respectively [15,17]. Studies that used decision tree modelling did not report time horizons as decision trees have no time-related component [16,18,19]. The two remaining studies did not report a discrete time horizon for the analysis [20,21]. Two publications reported using discount rates of 10% [15,21], while one publication reported using a discount rate of 3% [20]. The remaining publications did not mention any discounting [16–19], which was probably due to the fact that decision trees were used and therefore had no time horizon that or the time span modelled was one year or less. Two of the seven articles made explicit references to willingness-to-pay (WTP) thresholds for the cost-effectiveness of HAT as US$25/DALY [16,17]. One article mentioned that the WHO-CHOICE (CHOosing Interventions that are Cost-Effective) considered the gross domestic product (GDP) per capita of a country to be used as the WTP threshold for choosing between competing interventions [21,23]. The remaining publications [15,18–21] made no reference to a WTP threshold for the economic analysis under evaluation.

Base Case and Sensitivity Analyses

A full description of the economic outcomes for each study is outline in Table 5. The results from the sensitivity analyses conducted for the included publications are provided in Table 4.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. ICER results from economic evaluations.

https://doi.org/10.1371/journal.pntd.0003397.t005

A total of 5 studies [16,18–21] discussed cost-effectiveness results by calculating incremental cost-effectiveness ratios (ICERs), which are summarised in Table 5. Lutumba and colleagues published cost-effectiveness analyses of varying diagnostic algorithms for HAT [18,19]. Their results in 2005 demonstrated that lymph node puncture (LNP) in addition to CATT was more cost-effective ($20/LYS) relative to CATT alone or LNP alone [18]. In 2007, LNP followed by capillary tube centrifugation (CTC) and mini-anion exchange centrifugation technique (mAECT) (€76/LYS); LNP followed by thick blood film (TBF), CTC, and mAECT (€200/LYS); and LNP followed by TBF, CTC, mAECT, and CATT titration (€2,618/LYS) were deemed cost-effective relative to four other diagnostic algorithms. Although the strengths of these cost-effective algorithms were noted, Lutumba and colleagues noted that some of these algorithms may not be feasible to carry out in the field [19]. In regards to treatment regimens, Politi’s analysis [16] in 1995 demonstrated that based on a WTP of US$25/DALY, melarsoprol alone (initial treatment and relapses) was cost-effective at US$8/DALY (US$209/LYS) compared to no treatment. Politi’s analysis also demonstrated that a treatment pathway of melarsoprol with treatment relapses on Eflornithine (difluoro-methylornithine [DMFO]) (US$41/DALY and US$1,033/LYS) or DMFO for both treatment and relapses (US$167/DALY and US$4,444/LYS) would not have been considered cost-effective based on the aforementioned cost-effectiveness threshold of US$25/DALY [16]. A more recent publication by Robays demonstrated that DFMO was more cost-effective than melarsoprol (US$1,596/LYS and US$58/control case detected) when donated drug costs were not included; the analysis of cost-effectiveness was based on WHO-CHOICE’s suggestion that interventions at a cost of GDP per capita are very cost-effective and interventions at three times GDP per capita are cost-effective [24]. When donated drug costs were included, Robays found that DFMO was more cost-effective than melarsoprol at US$8,169/LYS and US$299/control case detected. Lutumba et al. [20] found that active screening (case detection) in addition to treatment was more cost-effective than treatment alone at $17/DALY averted and $301/control case detected or patient cured.

Two studies [15,17] did not report cost and effect results incrementally. Although Shaw et al. [15,17] conducted several analyses exploring combinations of case detection, diagnostics, treatment, and vector control, outcomes were not compared incrementally; consequently, ICERs were not attained. They did calculate $/patient detected with varying prevalence for five strategies and found that lower prevalence rates were associated with higher $/DALY and higher prevalence rates with lower $/DALY; these were based on average cost-effectiveness ratios, not ICERs.

All but one study included some form of one-way sensitivity analysis (OWSA). No studies completed subgroup analyses or conducted probabilistic sensitivity analyses (PSA), and hence, results were not presented using cost-effectiveness acceptability curves (CEAC). Additional measures of uncertainty were not explored in the form of a value of information (VOI) analysis in any of the reviewed publications.

Potential Use of Cost-Effective Modelling for HAT Control and Elimination

It was quite evident from the literature review that modelling will play a role in the economic evaluation of HAT. Most of the previous economic evaluations conducted were based on models, and modelling is known to assist with forecasting future economic consequences [13]. Decision makers would benefit from the use of whole disease modelling of alternative elimination scenarios because it would allow them to consider the implications and incremental benefits of each potential strategy. Previous economic evaluation studies reliant on modelling have addressed how individual interventions reduced transmission but not how these interventions, or combinations of them, could lead to eventual elimination or interruption of disease transmission. Current modelling techniques for economic evaluation, including those used to evaluate the impact of uncertainty related to model parameters, would also be useful for decision makers in communicating the consequences of choosing non-cost-effective strategies [36]. Additionally, modelling the feasibility of interventions through health service delivery is also necessary. For example, the results from an economic evaluation regarding diagnostic algorithms [19] showed that sometimes even the most cost-effective tools may not be affordable or feasible in some of the locations where HAT occurs [19].

Potential Use of Economic Evaluation Methodology in HAT Control and Elimination

A few considerations of cost-effective interventions could be gleaned from the few economic evaluations found. This was highlighted in the scenario described by Lutumba et al. [20] in which case-detection with treatment was more cost-effective than treatment alone, and an economic evaluation of diagnostic algorithms showed that the addition of tests to CATT could increase cost-effectiveness [19]. Treatment regimens including melarsoprol and eflornithine were considered cost-effective [16,21] for patients with HAT T. b. gambiense, and Politi’s analysis in 1995 also demonstrated a good understanding of economic outcomes because dominance was assessed and the importance of the efficiency frontier was illustrated [16]. Dominance refers to the economic concept that an intervention that costs less and has better outcomes relative to its comparator is considered dominant [13]. In regards to budgeting, sensitivity analyses [15,17,18] demonstrated that prevalence is related to costs. This will be important to consider because the cost per patient will increase towards the end goal of HAT elimination, but the overall cost per benefit still needs to be ascertained.

The economic evaluations reviewed presented some methodological inconsistencies. For example, there was a lack of clarity in reporting costs and consequences incrementally to a base-case scenario or relative to the next-best intervention. Historically calculations may have been done this way because of the “generalized cost-effectiveness” method [37], but if incremental and net benefits are always compared to “do nothing” instead of to the next-best option available, then the consequences of this methodology could lead to error [38]. Furthermore, when multiple strategies are being considered, dominance needs to be examined. Although four out of seven studies had more than two competing strategies, dominance was only addressed once. Evaluations that ignore dominance could lead to decision errors in which the health utility is not maximised at a societal level [13,39]. Cost-effectiveness was also referred to by the authors without making reference to a cost-effectiveness threshold. WHO-CHOICE [24] has defined thresholds previously; however, it is not clear if these thresholds values are acceptable for all global stakeholders because the authors did not always refer to a threshold value to determine cost-effectiveness.

The methodology of CEA with different interventions permits one to compare varying strategies across a disease, but the outcomes need to be unified so that decision makers can assess these comparators with ease and clarity. It is evident from this review that although CEA research may be conducted, the results are hard to interpret without standardization or reporting in a common metric (e.g., cost per DALY). Following existing guidelines for economic evaluation such as the SIGN Guidelines [14] and the more recent Consolidated Health Economic Evaluation Reporting Standards (CHEERS) statement [40] or developing guidelines that stakeholders feel acceptable for an elimination strategy would allow for consistency of analyses for HAT and other neglected tropical diseases. Formal economic evaluation guidelines and even a standard reference case have been developed by various public health funders [41–43], and researchers should consider these standards to further the future of CEA within tropical disease and disease elimination decision-making. In addition, traditional CEA measures two outcomes (cost and effects), but programs for elimination also need to consider time. Health economists will need to consider how to make recommendations to stakeholders for strategy prioritization considering all three elements for elimination.