Article Text
PDF
Abstract
Objectives More than 80% of cardiovascular diseases (CVD) and diabetes mellitus (DM) burden now lies in low and middle-income countries. Hence, there is an urgent need to identify and implement the most cost-effective interventions, particularly in the resource-constraint South Asian settings. Thus, we aimed to systematically review the cost-effectiveness of individual-level, group-level and population-level interventions to control CVD and DM in South Asia.
Methods We searched 14 electronic databases up to August 2016. The search strategy consisted of terms related to ‘economic evaluation’, ‘CVD’, ‘DM’ and ‘South Asia’. Per protocol two reviewers assessed the eligibility and methodological quality of studies using standard checklists, and extracted incremental cost-effectiveness ratios of interventions.
Results Of the 2949 identified studies, 42 met full inclusion criteria. Critical appraisal of studies revealed 15 excellent, 18 good and 9 poor quality studies. Most studies were from India (n=37), followed by Bangladesh (n=3), Pakistan (n=2) and Bhutan (n=1). The economic evaluations were based on observational studies (n=9), randomised trials (n=12) and decision models (n=21). Together, these studies evaluated 301 policy or clinical interventions or combination of both. We found a large number of interventions were cost-effective aimed at primordial prevention (tobacco taxation, salt reduction legislation, food labelling and food advertising regulation), and primary and secondary prevention (multidrug therapy for CVD in high-risk group, lifestyle modification and metformin treatment for diabetes prevention, and screening for diabetes complications every 2–5 years). Significant heterogeneity in analytical framework and outcome measures used in these studies restricted meta-analysis and direct ranking of the interventions by their degree of cost-effectiveness.
Conclusions The cost-effectiveness evidence for CVD and DM interventions in South Asia is growing, but most evidence is from India and limited to decision modelled outcomes. There is an urgent need for formal health technology assessment and policy evaluations in South Asia using local research data.
PROSPERO registration number CRD42013006479.
- cost-effectiveness analysis
- economic evaluation
- systematic review
- cardiovascular diseases
- diabetes mellitus
- South Asia
This is an Open Access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/
Statistics from Altmetric.com
- cost-effectiveness analysis
- economic evaluation
- systematic review
- cardiovascular diseases
- diabetes mellitus
- South Asia
Strengths and limitations of this study
This is the first systematic review to synthesise cost-effectiveness evidence on all types of interventions (policy, clinical or behavioural) to control cardiovascular diseases and diabetes mellitus in South Asia.
This review used a rigorous and broad search strategy including a wide range of sources to ensure all published studies are included for review.
This review used explicitly stated methods (protocol paper published) and standard checklists to assess methodological quality of studies.
The search was confined to English language publications performed as of August 2016, and this review excluded unpublished and ‘grey’ literature domain as we wanted to include studies that have undergone peer review process.
Significant heterogeneity in analytical framework and outcome measures used in these studies restricted meta-analysis and direct ranking of the interventions by their degree of cost-effectiveness.
Introduction
Evidence from randomised trials suggests that both pharmacological and non-pharmacological strategies are important in prevention and management of cardiovascular diseases (CVD) and diabetes mellitus (DM).1–12 While there is strong evidence on cost-effectiveness of pharmaceutical and lifestyle interventions in reducing the CVD and DM risk in affluent settings,13–16 little is known about the comparative cost-effectiveness of various interventions to control CVD and DM in South Asia. To generalise results from high-income countries to low and middle-income countries (LMICs) is not entirely justified because reasonable thresholds for cost-effectiveness will vary markedly—as will affordability. Also, setting specific cost-effectiveness information is important because of the differences in healthcare infrastructure.
With the rapidly increasing prevalence of CVD and DM in South Asia and the consequent huge economic losses, coupled with ill-equipped health systems and scarce resources to tackle the burden of chronic conditions, it is imperative to promote the most cost-effective interventions in this region. While a large number of economic evaluations have been recently performed in context to LMICs, and some authors have reviewed the available literature on non-communicable diseases broadly,17 18 no systematic attempt has been made so far to compile the evidence base and appraise the methodological quality of the economic evaluations of interventions to control CVD and DM in South Asia. To the best of our knowledge, no review has considered the cost-effectiveness evidence of interventions to control CVD and DM simultaneously, although these diseases share common risk factors.
We systematically reviewed the cost-effectiveness evidence on individual-level, group-level and population-level interventions to control CVD and DM in South Asia. The specific objectives were the following:
to summarise the incremental resource use, costs, consequences and cost-effectiveness of interventions versus comparators to control CVD and DM in South Asia
to describe the quality of economic evaluations considering key methodological issues.
Research design and methods
A protocol for the systematic review has been published previously and it provides a detailed description of the methodology, used for the current study.19 The systematic review has been registered previously in PROSPERO (CRD42013006479).
Briefly, we searched for studies that met the following inclusion criteria:
type of studies: full economic evaluations (cost-effectiveness analysis, cost-utility analysis, cost-benefit analysis) based on randomised trials or observational studies or decision models
type of participants: studies that included individuals with either established DM or CVD or at risk of developing these diseases in one of the South Asian countries: Afghanistan, Bangladesh, Bhutan, India, Pakistan, Maldives, Nepal and Sri Lanka
types of interventions: interventions or strategies for prevention and treatment of CVD or DM as documented in the previously published protocol19
types of outcome measures: we included several outcomes broadly under three domains—resource use, costs and cost-effectiveness as incremental cost per quality-adjusted life years (QALYs) gained, or disability-adjusted life years (DALYs) averted, or life years gained or intermediate outcomes; a detailed list has been presented in the previously published protocol19
studies published in the English language.
We searched 14 electronic databases and hand-searched for publications of the Disease Control Priorities Project 2 (DCPP2) and the WHO-Choosing Interventions that are Cost-Effective (WHO-CHOICE) to identify relevant studies. The details of the databases searched and a search strategy are provided in supplementary web appendix 1.
Supplementary file 1
Critical appraisal of included studies
Checklists proposed by Drummond et al,20 Evers et al 21 and Philips et al 22 were used for data extraction and to review methodological quality and strength of economic evidence. Also, we looked for funding sources of included studies.
Analysing, interpreting and reporting results
We extracted the incremental cost, incremental effect and incremental cost-effectiveness ratios (ICER) for interventions evaluated in the eligible studies. To adjust for cost and varying currencies over time, we used country-specific consumer price inflation rate to present value in 2017 and then used midyear currency conversion.23 24 All costs were converted to US$ (2017). Data extraction and critical appraisal of included studies were conducted by two authors independently and differences if any were resolved by consensus.
We used country-specific gross domestic product (GDP) per capita threshold, as per WHO guidelines,25 to interpret the ICER for all interventions evaluated in this review. We colour-coded ICER estimates as per the following scheme:
green=ICER<1×GDP per capita per QALY gained (highly cost-effective)
yellow=1–3×GDP per capita per QALY gained (cost-effective)
red=ICER>3×GDP per capita per QALY gained (not cost-effective).
Interventions that resulted in a negative incremental effect were regarded as dominated strategy and no ICER was reported. Further, we synthesised the cost-effectiveness data and presented the ICER for policy or clinical interventions, separately in the following categories: primordial, primary, secondary and tertiary prevention.
Difference between protocol and full review
We have not planned to include economic evaluations based on observational studies in the protocol but we have included it in our review. The more inclusive criteria enabled us to provide a more comprehensive review of the evidence base surrounding the topic. Risk of bias assessment in randomised trials was not conducted using Cochrane methods as Drummond and Evers checklists are inclusive of methodological quality assessments of economic evaluations alongside randomised trials as well.
Results
Search results
Our first search yielded 2949 items, titles and abstracts screening resulted in 85 articles, and full-text screening provided 42 articles that met full inclusion criteria (figure 1).
PRISMA flow chart for the selection of economic evaluations of interventions to control cardiovascular disease and diabetes mellitus in South Asia. CEA, cost-effectiveness analysis; CINAHL, Cumulative Index to Nursing and Allied Health Literature; CRD, Centre for Reviews and Dissemination; CVD, cardiovascular disease; DCPP2, Disease Control Priorities Project 2; EE, economic evaluation; HEED, Health Economic Evaluation Database; PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses; T2DM, type 2 diabetes mellitus; WHO-CHOICE, WHO-Choosing Interventions that are Cost-Effective.
Characteristics of included studies
Table 1 shows the detailed description of the studies (n=42) by country/setting, study population, intervention(s), comparator(s), economic perspective and type of analysis, and outcome measures.
Description of the economic evaluations and risk of bias assessment in the included studies
Study design
The economic evaluations were based on observational studies (n=9), randomised controlled trials (RCT) (n=12) and decision models (n=21).
Study setting
Most studies were from India (n=37), followed by Bangladesh (n=3), Pakistan (n=2) and Bhutan (n=1). Decision modelling studies had used effectiveness data mostly from meta-analysis of RCTs that reported results from developed countries.
Study population
Individuals (or population) at risk or with established CVD or DM were included.
Intervention targets and comparators
Three hundred and one interventions (policy, clinical or behavioural) were evaluated against null scenario (no intervention) or active comparators.
Perspective
In two-thirds of the studies (n=28), the authors explicitly documented and justified the economic perspective of the study. The studies used ‘health system’, that is, direct costs incurred by the health system (n=26); ‘patient’, that is, out-of-pocket payments by patient (n=6); or ‘societal’, that is, inclusive of all direct and indirect costs, plus productive loss (n=6) perspectives. Five studies did not state any perspective.
Funding
Two-thirds of evaluations (n=29) provided statements on the funding source. Public sponsorship or charitable trust/foundation grant was most common (n=16), followed by pharmaceutical industry (n=6) or received no support (n=7). A large number of studies did not state their source of research funding (n=13).
Resource use and costs
Only 20% of the studies (n=8) reported types and quantities of resource use and unit costs separately. Of these, five were RCT-based economic evaluations and two were decision model studies, suggesting that RCT provides an advantage on the reporting of actual resource use data as it is being collected during the trial.
Mostly direct medical costs were considered, although the scope of this varied enormously. For instance, 14 studies included only cost of intervention (medicines, diagnostics), while others (n=28) included cost of training, delivery of intervention, associated healthcare visit costs and travel cost of patients to the healthcare facility. Most (n=27) appeared to use an ‘ingredients’ costing approach, where costs were broken down between the main cost components such as medications, healthcare visits, vehicles, salaries and consumables. Fewer (n=5) used an ‘activity’-based approach, by identifying specific tasks such as programme and therapy costs. Two studies appeared to use some combination of the two, and it was not possible to discern the approach for eight papers. Few studies (n=6) also included ‘productivity losses’ (often termed ‘indirect costs’) in their assessment of costs, which were measured using the ‘human capital approach’.
Regardless of the approach taken, most papers (n=21) presented aggregated cost information. Many studies used actual expenditure data (n=17) as their source of costs data. Seven studies used published sources to generate cost estimates sometimes supplemented with expert opinion. Currencies reported were mostly in US$ (n=25), international dollars (n=4) or local currencies (Indian rupees/Bhutanese rupees) (n=6). In addition, seven studies quoted costs in both US$ and the local currency.
Outcome measures (consequences)
Nearly half of the studies (n=21) used ‘life years gained’ or ‘QALYs’ or ‘DALYs’ in their analysis. The calculation of QALYs/DALYs was based on South Asian population life expectancies; however, the utility values (QALYs weight) were derived from developed countries. Disability weights used in the WHO-CHOICE-based decision model studies were derived from the Global Burden of Disease (GBD) study (2000).26 The remaining studies reported intermediate outcome measures such as number needed to treat, length of hospital stay, hospitalisation rate, blood pressure (BP) reduction or CVD events avoided, which are easier to measure but harder to compare across interventions. None of the studies expressed outcomes (benefits) in monetary units.
Time horizon
Three-fourths of studies (n=31) explicitly stated their analytical time horizon. Eighty per cent of decision model studies adopted lifetime horizon and others reported cost-effectiveness estimates for 10, 20, 25, 30 or 50 years. RCT/observational studies-based economic evaluations had a median time horizon of 1 year.
Discounting
A discount rate of 3% was most often used for both costs and effects in decision model studies. RCT-based economic evaluations used a discount rate of 3% (n=3) and 5% (n=1). Further, 11 studies did not apply any discount rate.
Analytical approach
Cost-effectiveness analysis or cost-utility analysis were the main methods (n=34), followed by cost-consequences analysis (n=6) and cost-minimisation analysis (n=2). Although several of these papers (n=8) described themselves as cost-effectiveness analysis, they were in fact cost-consequences analysis or cost-minimisation analysis because an incremental analysis was not reported or there was no significant difference in the effectiveness of the intervention versus comparator, respectively. Most studies reported average cost-effectiveness ratio and interpreted it as ICER against the comparator as null scenario, that is, no intervention.
We found several different types of decision models used for cost-effectiveness analysis. A large majority of the studies used the WHO-CHOICE state transition model. Others used coronary heart disease (CHD) policy model, GeDiForCE, IMS Centre for Outcomes Research Diabetes Model, Centers for Disease Control and Prevention (CDC) model, Markov model or individual microsimulation model. Few studies provided details of model validation.
Sensitivity analyses and generalisability of study results
Nearly half of the studies (n=25) undertook some form of sensitivity analysis to assess the robustness of their findings to assumptions about input parameters. Of these, one-way sensitivity analysis was most often applied. Two studies used threshold analysis and one performed a multi-way sensitivity analysis. None considered the structural variations in the decision model for sensitivity analysis. Few studies described the model validation methods.
Three-quarters of the studies (n=32) discussed the generalisability issue. Efforts were largely confined to stating the limitations of the study, such as whether randomisation was employed or noting one or two facts about the study site which might limit generalisability to other contexts. Another 12 studies discussed issues of affordability but in brief terms, for example, by noting that the available budget should be taken into account (most studies focused on the cost-effectiveness without considering the budget impact/constraint) or by questioning the sustainability of a novel service such as a mobile diabetic retinopathy services, where there are already existing health services.27
Risk of bias assessment
In our critical review of methods used in economic evaluations to assess risk of bias, we found that almost all economic evaluations based on observational study only presented costs and consequences of two treatment strategies separately, without reporting an ICER or employed sensitivity analysis to assess robustness of costs or treatment effect estimates. Also, estimates of treatment effects from the observational studies are not very reliable due to the limitations in the original study design. On the other hand, economic evaluations based on RCTs reported better economic outcomes, that is, ICERs; however, these studies were limited by short follow-up duration (30 days to 1 or 2 years), treatment effects assessed as intermediate clinical outcomes (BP reduction, number needed to prevent one DM case) and mostly direct medical costs from health system perspective or patient perspective were reported, which ignores the societal costs and productivity loss due to illness. Lastly, decision modelling studies reported ICER per QALY gained or DALY averted mostly using the WHO-CHOICE methods, Markov models or microsimulation models from societal or health system perspectives. Many of the decision model studies from DCPP did not report the source of costs data, source of QALY weights and details on decision model structure or validation methods. Further, most of the WHO-CHOICE-based generalised cost-effectiveness analysis used disability weights from an earlier version of the GBD study (2000). Therefore, findings from this review should be used with caution for local decision making, and there is an urgent need for more investment in local research to generate evidence/data on costs of treatment and health services and effectiveness of interventions (table 1).
Methodological quality: summary
Figures 2 and 3 report the overall quality of studies based on the key methodological issues and technical characteristics for decision model studies, respectively. In general, very few studies reported quantities of resource use data and unit costs separately, details of statistical tests used and CI around ICER estimates. Among decision model studies, none reported methods used to assess methodological, structural or heterogeneity uncertainties, and very few discussed model validation methods. Critical appraisal of studies revealed that there were 15 excellent (++), 18 good (+) and 9 poor quality studies (−) (table 2).
Methodological quality of included studies. This figure presents the number of studies meeting the key methodological quality metrics of economic evaluations as recommended in the standard checklists.
Technical characteristics of decision modelling studies. This figure presents the number of decision modelling studies meeting the key methodological criteria for decision modelling studies as proposed by Philips et al. 22
Technical characteristics of included studies and quality grading (strength of evidence)
Cost-effectiveness evidence
Interventions reviewed for their cost-effectiveness are grouped under the scheme of primordial, primary, secondary and tertiary prevention of CVD and DM (table 3). This flow is used to make information available in an accessible format for policy-level and clinical decisions. Cost-effectiveness results from observational studies have not been included in the final synthesis of cost-effectiveness data from South Asia due to poor quality of evidence. Cost-effectiveness data presented below are for India unless otherwise specified (the GDP per capita (in US$ 2016) for India, Pakistan and Bhutan are 1861.5, 1468.2 and 729.5, respectively).28
Cost-effective interventions to control CVD and DM in South Asia
Primordial prevention
We found that a multicomponent population-level policy intervention consisting of increase in tobacco tax, clean indoor air law, advertisement ban and information/labelling are all highly cost-effective than increased tobacco tax alone (<1×GDP per capita per DALY averted).29 Addition of ‘nicotine replacement therapy’, ‘brief advice’ or ‘physician counselling’ to the combination strategy for tobacco control was not cost-effective (>3×GDP per capita per DALY averted).29 Complete smoking ban in public places is also highly cost-effective in terms of life years gained and acute myocardial infarction averted.30 School-based smoking prevention programme as evaluated in a cluster randomised trial in India31 was found to be cost-effective (1–3×GDP per capita per QALY gained). Salt reduction by legislation was cost-effective (1–3×GDP per capita per DALY averted).29 32 Substitution of trans fat with polyunsaturated fatty acids was cost-effective compared with null scenario (no intervention) per DALY averted.32 Media campaign to reduce saturated fat content was also cost-effective per DALY averted.32 A combined intervention of salt reduction by means of legislation together with public education campaign is cost-effective too.32 Alcohol taxation combined with advertisement ban was the most cost-effective strategy for alcohol control.15
Primary prevention
A 2015 modelling study conducted in Bhutan demonstrated that universal screening for diabetes and hypertension was highly cost-effective compared with no screening (<1×GDP per capita per QALY gained).33 Another 2006 modelling study from India34 showed that screening undiagnosed diabetes and treating those who test positive were not cost-effective, with an ICER of US$11 671 per DALY averted (ie, >3×GDP per capita for India), suggesting that screening for diabetes alone was not cost-effective and it should be supplemented with other risk factors, for example, hypertension. Other factors that could have influenced conflicting results include different health system-related cost, different model structure/model parameters, disease prevalence and time period.
Screening for gestational DM to prevent DM was also cost-effective compared with no screening.35
Among clinical interventions, preventive multidrug treatment provided to those at >35% cardiovascular risk vs 5% cardiovascular risk over 10 years was more cost-effective.29 Combined strategy of home health education plus trained general physician for hypertension management was highly cost-effective per DALY averted than individual strategies or no intervention in Pakistan.36
Lifestyle modification (weight reduction, increased activity and healthy diet) was most cost-effective for prevention of DM, followed by metformin alone and combination of lifestyle modification plus metformin (1–3×GDP per capita).37
Secondary and tertiary prevention
Policies to expand access of drugs for acute myocardial infarction prevention and treatment were cost-effective per DALY averted.38 Also, expansion of national insurance to cover secondary or tertiary prevention of CVD was most cost-effective per QALY gained compared with status quo.39 Clinical interventions for secondary prevention of CVD are mostly cost-effective per DALY averted.29 ECG-based doctor referral to cardiac care unit versus no ‘ECG use’ was cost-effective per QALY gained.40
Many strategies for DM treatment and secondary prevention of macrovascular and microvascular complications were found to be highly cost-effective or cost-effective. Examples of highly cost-effective interventions are glycaemic control in people with glycated haemoglobin (A1c) >9% with insulin, oral glucose-lowering drugs, diet and exercise, BP control in people with >165/95 mm Hg, and foot care in people with high risk of ulcers (<1×GDP per capita per DALY averted).34 Basal insulin treatment versus oral glucose-lowering drugs was highly cost-effective (<1×GDP per capita per QALY gained).41 Diabetic retinopathy screening every 2–5 years versus no screening was cost-effective (1–3×GDP per capita per QALY gained).27
Combination of primordial, primary, secondary and tertiary prevention
Multicomponent strategies of salt reduction through legislation (increase tax), health education, plus treatment of individuals at 35% cardiovascular risk with statin, diuretic, beta-blockers and aspirin were highly cost-effective, followed by similar strategy in those at 25% or 15% cardiovascular risk over 10 years.29 Policy measures such as expansion of insurance coverage for primary, secondary and tertiary prevention of CVD were also cost-effective (1–3×GDP per capita per DALY averted).39
Interventions that resulted in ICER>3×GDP per capita or were dominated by other highly cost-effective strategies are presented in online supplementary table 1. Significant heterogeneity in analytical framework and outcome measures used in these studies restricted meta-analysis and direct ranking of the interventions by their degree of cost-effectiveness.
Discussion
This review finds that, with some exceptions, most interventions to control CVD and DM were cost-effective (<1–3×GDP per capita per QALY gained or DALY averted), although the strength of evidence (and risk of bias) varied across economic evaluations based on observational studies, RCTs and decision models. Most interventions were cost-effective because of the large benefits in DALY averted or QALY gained at a marginal increase in cost per capita ($). These results should motivate decision makers to invest in primordial prevention strategies (increased tobacco tax, salt reduction by legislation, food labelling and food advertising regulation), and primary and secondary prevention interventions: multidrug therapy for CVD prevention and treatment in high-risk groups, lifestyle modification and metformin for diabetes prevention, and screening for diabetes complications every 2–5 years. Although detecting and treating diabetes earlier can prevent future complications and their associated medical costs, such savings were shown to be relatively small.34 An alternative to broad screening is to focus on targeted screening, that is, screening only persons with additional risk factors, such as hypertension and obesity. Such targeted screening was shown to be highly cost-effective or cost-saving when compared with no screening.33
Choice of comparator is an important decision when evaluating ICER of new interventions. In general, modelling studies that used the WHO-CHOICE method have reported average cost-effectiveness ratio against the null scenario (no intervention). In reality, however, this does not seem plausible because null scenario will not always reflect zero costs and zero effects. Also, these studies first identified the most cost-effective intervention among a group of strategies (eg, tobacco control, CVD prevention and treatment, or diabetes prevention and treatment) versus null scenario, then compared it with the next most cost-effective intervention.29 In many of such analysis, because the description of comparator was not clearly specified, the reported ICERs look ambiguous and changing the ‘comparator’ might produce a different ICER.
In our formal appraisal of the methodological quality of studies, we observed limitations in documentation of main study details, for example, chosen study perspective, sources of cost data and analytical time horizon. In addition, significant number of studies failed to provide details on units of resource use, costing year, currencies and other economic aspects. Since the discount rate used has an impact on cost-effectiveness estimates, the zero-discount rate applied in some studies is deeply concerning. In reality, however, every economic evaluation will contain some degree of uncertainty or imprecision. While one-way sensitivity analysis is helpful in understanding the impact of assumptions about one input parameter, multi-way sensitivity analysis offers a robust method to explore the uncertainty concerning more than one input parameters, but few studies reported results using this technique.
In terms of comparing results of this review with other contemporary reviews, we found cost-effectiveness evidence on a large number of preventive strategies, which is inconsistent with a previous review that examined the economic evaluation from Health Economic Evaluation Database42 and concluded that only 10% of all evaluations assessed preventive care. The greater number of preventive strategies found in our review could be due to the development of the WHO-CHOICE programme26 and the release of the DCPP2 in April 2006.43
Although cost-effectiveness evidence is available for 301 interventions to control CVD or DM, most of this evidence is based on decision models, which used data (annual risk of disease progression and intervention benefits) from Western countries. Most decision model studies have derived treatment effects from either meta-analysis of RCTs if available for an intervention or single RCT if meta-analysis is not available. However, the limited representation of South Asian populations in those RCTs remains an important concern. Therefore, our review highlights an alarming paucity of local research data to conduct high-quality economic evaluations and reflect the concerns of others in the field that large research gaps do remain in the area of health economic analysis in South Asian countries.44 Also, data from countries other than India are sparse. This is likely a reflection of research capacity in these countries, which needs to be addressed as a priority. Although the countries in South Asia are frequently grouped together, various countries in this region have substantially different health systems, health literacy, health indices, and hence healthcare needs. Understanding the differences be the countries is critical for policy makers, and therefore additional economic evaluations are urgently needed from other South Asian countries.
Strengths and limitations
This review has several strengths. This is the first study, to our knowledge, to include all types of interventions (policy, clinical and behavioural) that affect CVD or DM in South Asia. We considered all possible interventions (primordial, primary, secondary and tertiary prevention) to control CVD and DM together in this systematic review, primarily because policy makers have to choose between different options (competing priorities) for appropriate resource allocation, and as such a narrow economic research question is really not helpful for the systematic review, which intends to inform the process. We have used explicitly stated methods (protocol paper published)19 and standard checklists to assess methodological quality of studies. Recently, new methods have been proposed by researchers that can be applied to review decision model studies.45 However, use of new criteria would not change the findings of this review because these points have been covered broadly by the three popular checklists that we used in this review. Also, new methods have been proposed to estimate country-specific threshold for cost-effectiveness based on opportunity cost (health forgone) with investment in new intervention.46 But we preferred to present the findings based on WHO guidelines25 and for a lower threshold, that is, 1×GDP per capita. Moreover, the incremental cost and incremental benefits have been shown for all interventions (where available) so the decision makers or clinicians can make considerations based on their own willingness to pay threshold or budgetary constraints.
This review is not without limitations. First, the search was restricted to English-language publications performed as of August 2016. But this would not be a major problem because all the South Asian countries mostly publish research in English. Second, we excluded unpublished and ‘grey’ literature as we wanted to include studies that have undergone peer review process. We believe though that no major studies that can change the results of this review have been missed.
The review findings should be interpreted with caution because most of the cost-effectiveness studies were based on decision models. Although good-quality decision modelling study can provide information at a lower cost than RCT-based economic evaluations, models are based on assumptions and represent a simplification of—and therefore might depart from—reality. Furthermore, interventions that were highlighted as cost-effective (yellow) or highly cost-effective or dominant (green) analysed using the WHO-CHOICE framework could be reassessed by local agencies, particularly with regard to budget impact and also their cost-effectiveness, taking into account local costs and willingness to pay threshold value, similar to the work carried out by the Health Intervention and Technology Assessment Program in Thailand over the past decade.47
Future research directions
We have identified key research gaps in this review. Interventions involving multisectoral approach and policies for change in drug prices or devices (stents prices) have not been evaluated for their cost-effectiveness. The cost-effectiveness of these interventions should be assessed.
A few recommendations to advance the research on economic evaluations in the region are as follows. First, future studies need to take a broader societal perspective for analysis and present cost data in disaggregated form (resource consumption and unit costs, separately). Second, more research is needed to support the causes of variation among costs, effects and cost-effectiveness data on the universal screening of diabetes and/or hypertension. Third, research should focus on assessing the generalisability of cost-effectiveness analysis results within and between countries. Lastly, future cost-effectiveness analysis studies should adhere to international guidelines proposed by the WHO,25 International Society for Pharmacoeconomics and Outcomes Research,48–51 and the recommendations of the Second Panel on Cost-Effectiveness in Health and Medicine52 as a benchmark for design, conduct and reporting.
Conclusion
The existing economic evidence base from South Asia should motivate policy makers to mobilise resource allocation towards the most cost-effective interventions identified in this review to curb the epidemic of CVD and DM in the region. Also, there is an urgent need to invest in health technology assessment and policy evaluations in South Asia using local research data.
References
Footnotes
Contributors KS, NT, DP and AR conceptualised and designed the study. KS, AMCS and SB designed the search strategy for the review. KS and AMCS performed the search strategy in electronic databases, screened, reviewed and extracted data from eligible studies included in this review, and performed data analysis. KS wrote the first draft of the manuscript. AMCS, SB, KC, PDS, AUG, AR, DP and NT contributed significantly to the revision of the manuscript. All authors have approved the submission of this version of the manuscript.
Funding This research has received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests None declared.
Patient consent Not required.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement In this paper, we report the results of a systematic review. KS has access to all the data extracted from published studies. However, there are no unpublished data linked with this systematic review.