Introduction

High average systolic blood pressure (SBP) is responsible for a growing share of ill health in low-income and middle-income countries (LMICs), causing an estimated 12% of all deaths in 1990 and nearly 20% of all deaths in 2019.1 In 44 LMICs worldwide, only 3 in 10 adults with hypertension reported taking medication for this condition.2 Regional rates of control of high SBP ranged from fewer than 1 in 10 individuals with hypertension achieving control in sub-Saharan Africa to only 2 in 10 achieving control in Latin America and the Caribbean.2

Reducing the prevalence of uncontrolled hypertension is a long-standing global health objective.3 As the underlying cause of nearly half of cardiovascular disease (CVD) deaths,1 hypertension is a key condition whose treatment complements the 2030 Sustainable Development Goal (SDG) of reducing premature deaths from noncommunicable diseases (NCDs) by 25%.4 Yet, as countries seek to expand services for hypertension control, they may face resource limitations regarding the scale of their response.5

Treatment cut-points are blood pressure levels that serve as thresholds for initiating treatment. Applied at the population level, they can impact the number of people who receive treatment, and thus have implications for resource utilisation and health system capacity. Clinical practice guidelines (CPGs) that cast a wide net—treating all individuals at or above lower degrees of hypertension (eg, SBP≈140 millimeters of mercury (mm Hg)—generate the greatest overall good in terms of health outcomes, all else equal. However, in low-resource settings, such a wide-net approach may strain burdened health systems, increasing the risk that hypertension medications (often limited in availability) may not reach those with high severity SBP who would be expected to derive the greatest therapeutic benefit.6

Few economic evaluations have explicitly examined the health and economic consequences of different SBP treatment cut-points, nor examined how these might vary across countries of different income levels. As countries strive to meet the United Nations SDG 3.8 of achieving universal health coverage, each starting from a different baseline of health system resources, a deeper understanding of relative cost–benefits might inform efficient ways to achieve their goals.

This economic evaluation contributes to that understanding by analysing the costs and consequences of scaling hypertension treatment in 24 LMICs representing different income levels and regions. We seek to answer: what are the costs of scaling pharmacological antihypertensive treatment at three distinct cut-points, as well as the expected health (avoided CVD events, deaths, disability-adjusted life-years (DALYs)) and economic benefits (healthcare savings, monetised benefits of reduced mortality and morbidity). Evidence from the analysis can inform approaches for addressing unmet hypertension treatment needs in diverse contexts.

MethodsOverview

We built an Excel-based probabilistic state-transition model—online supplemental appendix 1—to calculate the costs and consequences of strengthening the care cascade compared with leaving it as it is today, while investigating the extent to which different SBP treatment cut-points affect outcomes. We conducted the analysis from a societal perspective to account for health and non-health sector consequences of improved health. The model assessed outcomes for adults aged 30+ years in 24 LMICs (Africa—Algeria (LMIC), Kenya (LMIC), Tanzania (LMIC), Uganda (Low income (LIC)). Americas—Argentina (Upper-middle income (UMIC)), Brazil (UMIC), Colombia (UMIC), Mexico (UMIC). Eastern Mediterranean—Iraq (UMIC), Libya (UMIC), Morocco (LMIC), Sudan (LIC). Europe—Azerbaijan (UMIC), Belarus (UMIC), Kyrgyzstan (LMIC), Tajikistan (LIC). South-East Asia—Bangladesh (LMIC), Myanmar (LMIC), Nepal (LMIC), Sri Lanka (LMIC). Western Pacific—(Cambodia (LMIC), Lao (LMIC), Mongolia (LMIC), Vietnam (LMIC)).

The 2 lower-income, 14 lower-middle-income and 8 upper-middle income countries were chosen based on the availability of model input data from recent WHO STEPS surveys, population size and representation of all WHO regions. We selected analysis countries from a list of LMICs that had conducted WHO STEPS (STEPwise Approach to NCD Risk Factor Surveillance) surveys between 2009 and 2019, choosing the four most populous countries in each region with STEPS data. For Latin America and the Caribbean—where no large countries had conducted STEPS surveys—analysts selected the region’s four most populous countries. A 30-year time horizon was selected to assess outcomes in the medium term. Costs of the interventions and economic benefits were reported in 2020 US dollars (USD), using a 4% discount rate that is recommended for economic evaluations of health programmes in middle-income countries.7

In recognition that ‘conducting both cost-effectiveness (CE) analyses and benefit cost-analyses (BCA) provides useful insights in many settings’,8 we compared intervention costs to DALYs gained to assess cost-effectiveness using 1× gross domestic product (GDP) per capita and two CE thresholds estimated by Ochalek et al—country income-group-specific thresholds9 and country-specific health opportunity thresholds.10 ,9The income-group specific thresholds are 0.18x, 0.15x and 0.55x GDP per capita for low-income, lower-middle-income and upper-middle-income group countries, respectively. The country-specific thresholds are Ochalek et al’s10 ‘DALY 3’ estimates, which we updated by applying the cost per DALY averted as a per cent of GDP per capita to each country’s 2020 GDP per capita (see our online supplemental appendix table A2). Both threshold types from Ochalek et al are opportunity costs, that is, they reflect the health benefits (in DALYs) that could be achieved from general investments in increasing health system expenditures as opposed to investments in specific interventions. We also compared intervention costs to monetised health benefits in the tradition of a BCA.

Online supplemental appendix 2 contains more information on referenced data and sources (online supplemental appendix tables A1–A12 and online supplemental appendix figure A1–A3), HEARTS protocols (online supplemental appendix figure A4), detailed results (online supplemental appendix tables A13–A18; online supplemental appendix figure A5–A7), a sensitivity analysis (online supplemental appendix figure A8) and model validity (online supplemental appendix figure A9).11

Model structure and analysed scenarios

The model simulated how mean SBP in the adult population would change based on changes in rates of hypertension screening, treatment and control (ie, the hypertension care cascade), and then assessed the resulting impact on CVD outcomes over time (figure 1). The primary hypertension measure in the model was SBP, based on evidence that SBP and diastolic blood pressure are correlated, and SBP is better associated with health outcomes.12 13

Figure 1

Hypertension model schematic. CVD, cardiovascular disease; SBP, systolic blood pressure.

The product of initial rates of screening, antihypertensive drug treatment and control is the ‘effective coverage’ of hypertension interventions—or the per cent of all individuals with hypertension who achieve control. After populating the model with data on countries’ hypertension care cascades, we considered the following scenarios.

In a status quo scenario, the hypertension care cascade is static over the 30-year time horizon of the analysis. This business-as-usual approach, in which effective coverage of hypertension interventions does not change, acts as a baseline against which to compare the intervention scenarios.

In three intervention scenarios, we simulated strengthening the hypertension care cascade so that more people are screened, receive treatment and control their blood pressure—with one important difference between the scenarios: the applied treatment cut-point.

Treat ≥160 mm Hg—In scenario 1, only adults diagnosed with severe levels of hypertension (SBP≥160 mm Hg) received treatment.

Treat ≥150 mm Hg—In scenario 2, only adults diagnosed with SBP≥150 mm Hg received treatment.

Treat ≥140 mm Hg—In scenario 3, adults with diagnosed SBP≥140 mm Hg received treatment.

These intervention scenarios allowed us to examine what would happen if countries’ ‘forward progress’ on hypertension care was directed at subsets of persons living with hypertension.

Illustratively, we assumed the following care cascade strengthening in all of the intervention scenarios: (1) the screening gap (ie, gap between initial screening rates and 100% screening rates) linearly decreased by 50% over 10 years; (2) the treatment gap (ie, gap between rates of diagnosis vs 100% treatment rates) also linearly decreased by 50% and (3) control rates (per cent of those treated who achieve control) were assumed to linearly climb to 60%, 61% and 71% in low-income, lower-middle-income and upper-middle-income countries. These represented ambitious advancements, with screening and treatment gaps closing at roughly double the worldwide rate from 1990 to 2015, and control rates representing the highest rate achievements to date in respective income settings.14 However, the resulting effective coverage rates—averaging 41% across 24 countries (see online supplemental appendix 2, table A4)—fall well within 80-80-80 targets for screening, treatment and control (51% effective coverage rates) that have been advocated as targets for hypertension.15

Comparing the effective coverage rates in the status quo and intervention scenarios, the model calculated the difference in two leading causes of hypertension-attributable ill-health—stroke and myocardial infarction,16 including (1) incidence of acute events, and attributable (2) deaths and (3) DALYs. We also quantified averted direct costs (ie, healthcare expenditures to treat acute and chronic CVD) and averted indirect economic costs (ie, morbidity-related productivity losses due to absenteeism and disability, and mortality-related losses).

Model epidemiology

Country-specific annual CVD incidence, prevalence and cause-specific and other-cause death rates were from the Institute for Health Metrics and Evaluation’s (IHME) Epi Visualisation database.17 We weighted measures to create rates for adults aged ≥30 years, by sex. Disability weights for health states are from IHME’s Global Burden of Disease (GBD) study.18

To begin, adult populations were stratified by existing health status—CVD-free, and those with history of stroke or ischaemic heart disease (IHD). The CVD-free population was further categorised by SBP level—140 to <150 mm Hg, 150 to <160 mm Hg and≥160 mm Hg—based on prevalence data from country surveys19–39 and the NCD Risk Factor Collaboration dataset.40 All adults in a cohort were assigned the mean SBP of the applicable SBP severity level.

IHME’s CVD-incidence rates are averages among the population. We needed blood pressure-specific CVD incidence rates for each distinct blood pressure severity level (ie, SBP 140 to <150 mm Hg). Using data on mean SBP in the population, and evidence on the relative risk of stroke and IHD events (from Ettehad et al41: every 10 mm Hg reduction in SBP reduces the relative risk of IHD events by 0.80 (95% CI 0.77 to 0.83) and for strokes by 0.73 (95% CI 0.68 to 0.77). Given that the authors did not differentiate between ischaemic and haemorrhagic stroke, we assume the same relative risk within the analysis),41 we calculated incidence rates for acute CVD events for each country, sex and blood pressure severity level. In the intervention scenarios, these rates reduce according to the SBP-lowering impact of medications (see below) and the proportion of the population achieving control—with the model also assessing changes in downstream health and economic outcomes because of the blood pressure lowering. The model also considers dynamic population changes in the target population (age 30+) according to resulting birth and death rates.

Blood pressure-lowering impact

We applied the SBP-lowering effects of various medication classes to treatment approaches outlined in WHO HEARTS evidence-based treatment protocols (see online supplemental appendix 2, figure A5).

Medication impacts were sourced from Wald et al42 who reported that the SBP-lowering impact of monotherapy was 6.8 mm Hg for ACE inhibitors (ACE-I), 7.3 mm Hg for thiazide diuretics and 8.4 mm Hg for calcium channel blockers (CCB).42 The ‘blood pressure-lowering effect of a drug is approximately 1 mm Hg less for each 10 mm Hg decrement in pretreatment blood pressure’ and adding a new drug is about five times as effective as doubling the dose of a drug.42

Using this rule, we took Wald et al reported blood pressure-lowering impacts by medication class and adjusted them for the starting average SBP level in each country. From table 1 in Wald et al,42 we calculated that the weighted mean SBP in trials assessing the impact of CCBs was 155.7 mm Hg, diuretics 150.5 mm Hg and ACE-I 156.2 mm Hg. Since a medication’s expected effect size changes based on a patient’s level of SBP, we adjusted effect sizes to estimate SBP-lowering effects at different hypertension severity levels. For example, while a diuretic lowers SBP 7.3 mm Hg in an individual with starting blood pressure of 150.5 mm Hg, it is only expected to lower SBP by 6.8 mm Hg in an individual with pretreatment SBP of 145: Adjusted impact=7.3 mm Hg−((150.5 mm Hg–145 mm Hg)×0.1)≈6.8 mm Hg). Then, starting from the initial impact of one drug, we calculated (see formula 1) the effect of combination therapy (two-drug or three-drug combinations) for each step of WHO Global HEARTS protocols.43 For example, after initial application of a diuretic expected to reduce SBP by 6.6 units in a person with a starting initial blood pressure level of 150, addition of a second medication expected to reduce blood pressure by 5.5 in a person with the same starting blood pressure level would lead to a total drop of 11.4 units (6.6+5.5–5.5×0.1).

Formula 1: Calculating the impact of antihypertensive medications administered in triple combination

Effect—the reported blood pressure-lowering effect (mm Hg) of a given drug as monotherapy.

i—specified pretreatment SBP level.

Among the 24 countries in our analysis, the mean SBP with the established blood pressure severity groups was, respectively, 144, 154 and 172 mm Hg. Results from our calculations suggested that monotherapy would be sufficient to meet standard ‘control’ goals (<140 mm Hg) in persons with SBP 140 to <150 mm Hg. Dual-combination therapy was sufficient for adults SBP 150 to <160 mm Hg, and triple-combination therapy was required to reduce those with SBP≥160 close to, but not below, the standard SBP control threshold.

Financial costs of interventions

We included two types of direct care costs: outpatient clinic visits and medications.

Reflecting on 10-year CVD risk levels at differing levels of SBP and following the WHO HEARTS Risk-based CVD Management technical guide’s recommendations on annual outpatient visits by CVD risk level, we determined 1.3 clinic visits are required annually for people with SBP 140–160 and 2.5 visits for those with SBP 160+.44 The country-specific cost of these visits is given by modelled estimates from WHO-CHOICE.45 46 The estimates are the ‘hotel cost’—all costs excepting medications and diagnostics—of a 10 min visit (The analysis assumes that the mean duration of a clinic visit is 10 min, following the WHO NCD Costing Tool) to a health facility without beds.

To obtain medication costs, we searched international databases47 48 and national documents and databases (A list of national medication databases was obtained from WHO partners (see the Acknowledgements section))49–61 and augmented our search via personal correspondence62 and other sources.63 64 We limited our search to public procurement prices to better reflect the economic cost of medications and to enhance comparability.

As we did not obtain country-specific public procurement prices for all analysed countries, we calculated the median public procurement price among countries with price information and then calculated the cost of each HEARTS first-line treatment protocol (see online supplemental appendix 2, tables A10–A12) initiating treatment with a single medication—since few countries have adopted combination therapy as a form of first-line treatment.65 Each protocol initiates treatment with a different medicine class (eg, CCBs), advancing to higher strengths and/or multiple medications using single-agent pills if hypertension persists. We determined that the HEARTS protocol using diuretics as a form of first-line treatment—and employing hydrochlorothiazide, amlodipine and losartan—was the least costly option. In the analysis, in each country, the diuretics protocol was applied using these three medications.

Other costs: screening, supply chain and programme costs

Other included costs are those to screen adults for high SBP; supply-chain costs to import, store and distribute medications and programmatic costs.

Screening for high SBP was assumed to occur within the health system during presentation at health clinic visits. Based on time and motion study estimates of provider time to ask about a patient’s health history, provide a physical exam (including SBP measurement) and counsel patients and document results, we assumed an additional 10 min of time was required in a health visit to screen for SBP.66–68 Using assumptions embedded in the WHO NCD costing tool, we determined the cost of those ten minutes added 21% to the cost of an outpatient visit. The administration of other diagnostic tests (eg, urine sample, lipid panels) was not costed.

Resources are required to procure, import, store and distribute medications. A USAID study reported that supply chain costs to deliver essential medicines—as a percentage of the overall cost of the medication—generally ranged between 12% and 20% in more-developed countries and between 20% and 25% in less-developed and geographically challenging countries, and can be as high as 35% in post-conflict settings and 50% in failed states.69 Our analysis adopted and adapted these estimates.

An updated70 WHO NCD Costing Tool71 provided NCD programme costs, including country-specific costs of human resources, advocacy, monitoring, training (for programme staff), meetings, mass media, supplies and equipment and other costs required to run a holistic NCD programme. Given that our analysis focused only on CVD—whereas a holistic NCD programme with a 5×5 strategy (Diseases—CVD, diabetes, cancer, respiratory diseases, mental health conditions. Behavioural risk factors—tobacco, alcohol, unhealthy diet, physical inactivity and pollution) would focus on multiple NCDs and NCD risk factors—we apportioned one-fifth of the Tool’s reported programme costs.

Economic value of improvements in health

We calculated the economic value of improvements in health. The value of these improvements in health is reported as ‘total economic benefits’—consisting of the value of reductions in CVD-attributable mortality, averted healthcare expenditures and averted indirect productivity losses.

Value of reductions in fatal health outcomes (CVD-attributable mortality)

Few studies have produced data required to estimate the value of mortality risk reductions in LMICs.72 Following Reference Case Guidelines for Benefit-Cost Analysis in Global Health, we extrapolated population-average ‘value of a statistical life-year (VSLY)’ for each analysis country from a 2020 US estimate,73 using gross national income per capita as the recommended proxy for income in comparison to the USA and an income elasticity of 1.5.8

Within the analysis, the model tracks adult mortality, and the number of years adults would have lived if they hadn’t experienced CVD-attributable mortality. Each additional year of life is valued at the country VSLY, adjusted for expected real income growth year-over-year.

Value of reductions in nonfatal health outcomes

With little available data to value nonfatal risk reductions in LMICs, we follow recommendations to use the value of averted healthcare expenditures and indirect productivity losses as a proxy.8

Averted health expenditures include those to treat acute cases and chronic cases of CVD. Country-specific acute IHD and stroke expenditures were estimated following Ding et al74 who used European Union health-expenditure data to generate estimates of CVD-event costs around the world.75 The model tracks stroke and IHD events averted as a result of treatment for hypertension. Each averted event was multiplied by the derived CVD-event cost.

We also estimated country-specific expenditures to treat chronic patients with CVD. We assumed four required primary care outpatient visits, aligned with Global HEARTS protocols for patients with high levels of CVD risk,44 with the cost of each visit given by previously described WHO CHOICE estimates. We applied medication costs to pharmacological treatment strategies given by HEARTS evidence-based treatment protocols for patients with chronic IHD or stroke.43 The model tracks CVD-event survivors and applies the calculated treatment costs, assuming that only a portion of adults with chronic CVD receive pharmacological treatment, according to rates described in STEPS surveys.

The analysis valued averted indirect productivity due to disability-induced labor-force drop-out and due to temporary work absenteeism due to ill health. To estimate labour force drop-out, we reviewed IHME GBD disability weights18 to identify poststroke and post-IHD events that could reasonably be categorised as significantly ‘disabling’ (We assumed sequela levels 3–5 are disabling for chronic intracerebral haemorrhage or chronic ischaemic stroke, and, for IHD, that sequala level three or ‘severe’ heart failure or severe angina is disabling). Next, we applied expected distributions of people into these disability states after stroke or heart attack events to calculate the number of people who experience a disabling event76—about 1 in 5 persons disabled after a stroke and 1 in 15 persons after an acute myocardial infarction. Finally, we accessed data on employment rates among individuals with and without a disability from the International Labour Organization77 78 to estimate how many people would drop out of the workforce due to disability. We multiplied that number by country-specific and sex-specific earnings for each year a person would have worked through average retirement age if they had not dropped out of the labour force—accounting for real wage growth and factoring in state-and sex-specific background mortality.

Two types of temporary workplace absenteeism include (1) days of work lost due to an acute CVD event and recovery and (2) days of work lost due to chronic CVD in years following an acute event. In the analysis, acute myocardial infarctions and strokes, respectively, caused adults to miss 39 and 57 days of work.79 Individuals with chronic CVD missed an additional 2.7 days of work per year.80 Averted absenteeism costs were calculated as the number of employed adults in a health state multiplied by days of work lost due to the health condition multiplied by annual country-specific and sex-specific earnings.

Patient and public involvement

None.

ResultsFinancial costs of the interventions

The financial costs—inclusive of screening, treatment, supply chain and programmatic costs—of treating to three different SBP cut-points in LMIC settings are described in figure 2. In the examined scenarios, a strategy that employed a treatment cut-point at SBP 140 mm Hg resulted in the highest costs, reflective of a decision to treat the highest number of people—that is, all individuals above the traditional cut-off for hypertension. With fewer people to treat, more selective treatment decisions lowered overall costs—by 18% at 150 mm Hg and 38% at a 160 mm Hg cut-point. On average across two low-income, 14 lower-middle-income and eight upper-middle-income countries, costs to fully fund scale-up of hypertension care using a 140 mm Hg cut-point, respectively, required 10.2%, 4.5% and 1.2% increases in annual government health expenditures. On average across countries, the highest share of costs was attributable to medications (54.8%), followed by outpatient clinic visits (26.5%), supply chain expenditures (10.8%), screening (7.2%) and programmatic costs (0.7%) (see also online supplemental appendix 2, tables A13 and online supplemental appendix, figure A6).

Figure 2

30-year financial costs by scenario (USD millions, discounted), the required per cent increase in current government health expenditures to fully fund scale-up of hypertension care, and financial costs by source. Financial costs by source are represented for the scenario using a treatment cut-point of 140+. See online supplemental appendix 2 for underlying data. Legend: Red = screening, Gray = medications; Blue = supply chain; Gold = clinic visits; Turquoise = program costs.

Improvements in health and their economic value

Over a 30-year time horizon in the 24 analysis countries, in the examined scenarios treating at a cut point of ≥SBP 140 mm Hg avoided 2.6 million CVD events, 1.2 million CVD-attributable deaths (7% of expected CVD deaths among persons with hypertension 140+) and 10.2 million DALYs.

Most (68%) of these health benefits were obtained from treating those with SBP≥160 mm Hg, despite that this severity group represents only one-third of persons living with hypertension (see online supplemental appendix 2, figure A2). Figure 3 shows that treating only adults with SBP≥160 mm Hg would save 806 000 lives. Expanding treatment to also include those with SBP≥150 to <160 mm Hg would save about an additional 227 000 thousand lives, and further expanding treatment to adults with SBP≥140 to <150 mm Hg would save 159 000 thousand more lives (see online supplemental appendix 2, table A14).

Figure 3

Averted deaths by treating specific SBP severity levels, total across all countries. See online supplemental appendix 2 for breakdowns of averted deaths by treatment cut-point by country. SBP, systolic blood pressure.

Figure 4 reports total 30-year economic benefits—inclusive of averted healthcare expenditures, averted indirect productivity losses and the value of averted mortality—by treatment cut-point. While overall, the highest economic benefits are gained from inclusively treating to a 140 mm Hg cut-point, nearly 7 out of every 10 dollars in economic benefits are generated because of treatment of the severe SBP group (160 mm Hg+).

Figure 4

30-year economic benefits (USD millions, discounted) and economic benefits by source. Economic benefits by source represented for the scenario using a treatment cut-point of 140+. See online supplemental appendix 2 for underlying data. Legend: Red = Averted healthcare expenditures; Gray = value of averted disability-induced labor force dropout; Blue = value of averted absenteeism; Gold = value of averted mortality.

By source, the highest share of economic benefits is the value of averted mortality (85.4%), followed by averted labour force drop-out (6.6%), averted healthcare expenditures (5.4%) and averted absenteeism (2.6%) (see also online supplemental appendix 2, table A15 and online supplemental appendix, figure A7). On average across countries, averted healthcare expenditures are equivalent to nearly 8% of the financial costs to scale hypertension care—ranging from less than 1% in Uganda to 27% in Brazil (see online supplemental appendix 2, table A16).

Net economic benefits by treatment cut-point

Table 1 lists countries from lowest to highest GDP per capita and displays the cost-effectiveness of hypertension treatment at the 140 mm Hg cut-point, as well as the net economic benefits for all three cut-points.

Table 1

30-year cost-effectiveness (CE), net benefits and treatment decisions, by treatment cut-point

Net economic benefits were calculated by subtracting the financial costs from the economic benefits associated with each treatment cut-point strategy. In general, net economic benefits were higher at higher country income levels, with 10 out of 12 countries projecting positive net benefits in at least one scenario compared with only 3 out of 12 for the lowest income countries.

In 9 out of the 12 countries in the bottom half by income, the scenario using a treatment cut-point of SBP≥160 mm Hg projected the largest net economic benefits. Among the countries in the top half by income, net economic benefits were maximised using a cut-point of SBP≥140 in six countries, SBP≥150 mm Hg in four countries and SBP≥160 mm Hg in two countries.

Although the estimated net economic benefits for some countries are larger at higher treatment cut-points (eg, 160+ mm Hg), additional deaths can nonetheless be averted by expanding treatment to lower cut-points (eg, 140+ mm Hg). The last column in table 1 represents the potential gains in lives saved—over 126 000—over 30 years by expanding treatment cut-points from 160 to 140 mm Hg regardless of net economic considerations.

Respectively using a country income-group-specific cost-effectiveness threshold,9 country-specific health opportunity CE threshold,10 and a 1×GDP per capita CE threshold, treating at the 140 mm Hg cut-point was cost-effective in 4 out of 24 countries, 7 out of 20 countries for which threshold estimates were available and 13 out of 24 countries. 67%, 70% and 88% of CE threshold decisions respectively agreed with net benefit decisions (ie, ‘not CE’ matched with negative net benefits at a 140 mmHg cut-point, or ‘CE’ matched with positive net benefits).

Discussion

Our analysis of scaling hypertension care in 24 countries—across a range of income levels and regions—estimated that compared with continuation of the status quo improving the hypertension care cascade using an SBP≥140 mm Hg cut-point would save nearly 1.2 million lives over the study’s 30-year time horizon. It also identified that around 68% of those health benefits are derived from treating those at the highest risk (SBP≥160 mm Hg).

Treating at higher cut-points (ie, ≥150 or ≥160 mm Hg) must be weighed against a range of quantitative and qualitative criteria. Clinical guidelines in LMICs—which vary significantly in recommended cut-points65—rarely detail how consideration of health, economic, equity and other criteria impact choices. More explicit frameworks for decision-making would benefit national health authorities. Once criteria are established, country-specific policy analyses that couple economic evaluations with assessment of qualitative criteria may assist national health authorities to hone-in on—and defend—choices to treat at specific cut-points.

This study provides some economic evidence to inform decision-making and prompt additional investigation. Considering only net economic benefits, countries with relatively higher incomes would generally benefit from treating at 140 or 150 mm Hg cut-points, while countries with lower incomes may maximise net economic benefits at the 160 mm Hg cut-point. In lower-income countries, the marginal difference in cost-effectiveness between treating at 160 and 140 cut-points is also higher than in upper-middle-income countries (see online supplemental appendix 2, table A18). These findings are in line with limited other studies that have examined the cost-effectiveness of hypertension treatment cut-points globally. A study of the cost-effectiveness of new recommendations to treat to a 130 mm Hg cut-point in the USA found that treating adults to below 140 mm Hg was generally not cost-effective, except in older adults and/or those with high CVD risk.81 82 Another in the UK concluded that pharmacological treatment for stage 1 hypertension (SBP 140 to <160 mm Hg) was cost-effective for most demographic groups but that cost-effectiveness diminished with treatment of lower CVD risk groups.83

Several factors—prevalence of other CVD risk factors, underlying epidemiology, existing treatment rates and local economic conditions—may contribute to hypertension care being generally more cost-effective if applied to those most at risk in lower-income countries. In some lower-income countries, populations may have lower levels of risk-factor comorbidity,84 which reduces overall CVD risk and thus the potential gains from lowering SBP.85 In addition, limited health system infrastructure and lower treatment rates for CVD in lower-income countries86 bound the potential of secondary prevention to avert downstream healthcare costs of acute CVD care. Finally, VSLY estimates, employment rates and earnings also increase with country income, even as major cost drivers such as medications may remain relatively similar across income strata or even decrease as a country’s purchasing power and income increase. Thus, in modelling exercises, higher-income countries may ‘outpace’ costs more easily as greater economic benefits accrue as rewards to improved health. Our sensitivity analysis demonstrates that a country’s VSLY and earnings have material impacts on assessed outcomes (see online supplemental appendix 2, table A17). Together, these factors contribute to the potential differences in net economic benefits across income strata.

Focusing only on economic criteria presents a moral dilemma given the loss of a myriad of lives that otherwise could be saved by treating to a lower cut-point. But, in some circumstances, budget, available human resources, and health system infrastructure may suggest that scarce resources are triaged to those at the highest-risk first—while preparing the health system to expand to meet the needs of all patients with hypertension. This underlines the essential need to reach those most at risk and is supportive of continued expansion of programmes—such as WHO HEARTS—that aim to strengthen health systems’ capacity to assess total CVD risk. With assessment, health systems may be able to target additional resources and interventions to increase the likelihood of treatment adherence in high-risk individuals.

In our study, scaling hypertension care using an inclusive 140 mm Hg cut-point required high increases in government health expenditures in low-income and lower-middle-income countries: 10.2% and 4.5%. Using a more restrictive 160 mm Hg cut-point would cost about 40% less. Regardless of country income status, there are opportunities to increase the cost-efficiency of treatment. More data is needed on medication costs—about half of the costs of scaling hypertension care in our study. As demonstrated in this study’s sensitivity analysis (see online supplemental appendix 2, figure A8), medication costs were one of the top-five factors impacting cost-effectiveness, giving impetus to the need to tackle the affordability. National governments should leverage pooled procurement mechanisms and increase transparency around medications prices.87 Opportunities may exist to refine national and international guidelines to promote cost-effectiveness. In subanalyses, this study investigated cost-minimising treatment regimens (see online supplemental appendix 2, table A12) and the cost-efficiency of combining antihypertensive medications within a patient’s treatment regimen versus increasing dosages of medications that patients are already taking (see online supplemental appendix 2, figure A5)—each of which could be considered in development of new CPGs.

Limitations

The model does not consider comorbid risk factors or diseases (eg, chronic renal disease)—despite that hypertension treatment decisions are often guided by overall CVD risk. Depending on the population prevalence of risk factor comorbidity, high-risk populations would benefit more from SBP-lowering than the model predicts. Explicit consideration of overall CVD risk would add clarity to the merits of treating at different risk cut-points, especially given that population risk profiles evolve as country incomes increase.

Our analysis captures leading causes of hypertension-attributable death—ischaemic and haemorrhagic stroke and myocardial infarction16—but it does not capture other diseases linked to hypertension, including congestive heart failure, coronary artery disease,88 cardiomyopathy, atrial fibrillation, aortic aneurysm, rheumatic heart disease, peripheral vascular disease, endocarditis, chronic kidney disease,16 dementia and Alzheimer’s.89 Other downstream impacts—such as averted presenteeism or caregiver time—were not considered in the analysis. In addition, we assumed that a certain proportion of the population with hypertension achieved control; but those who did not achieve control were assumed not to have any blood-pressure lowering effect. Together, these factors contribute to a more conservative estimate of the benefits of lowering SBP.

National-level data for all countries in this analysis were not always available. One important example is medication costs, which accounted for about half of all financial costs in the analysis. Given that a country’s relative size and purchasing power may drive public procurement price and that medication costs have previously been found to vary considerably across countries,90 including by income level,90 the medication prices in our analysis may not be representative of those countries are able to negotiate. Moreover, we also assumed the deployment of cost-effective medications and protocols, which may underestimate costs. Finally, given insufficient evidence on labor productivity impacts of CVD in LMICs, model inputs such as disability rates and absenteeism—were largely derived from high income settings.

In our scenarios the care cascade—screening, treatment and control rates—strengthened over time. We assumed progress given that, within our scenarios, governments fully covered costs—erasing affordability concerns—and we accounted for supply chain costs that would improve availability. The resulting effective coverage rates are within realistic implementation goals.15 But other barriers—for example, patient social support, health literacy, acceptability of treatment—influence the hypertension care cascade91 and additional intervention (eg, mass-media awareness campaigns, fixed-dose combination medications to reduce pill burden) will likely be needed to achieve high effective coverage. Without holistic attention to barriers at the patient and provider levels, countries will continue to invest resources that do not ‘pay off’ given fall off along the cascade of care. The sensitivity analysis shows that failing to achieve control has the highest impact on cost-effectiveness of hypertension care (see online supplemental appendix 2, figure A8).