Benefits, costs and efficiency indices

In a previous work, we programmed a computing algorithm that: 1) selects population individuals having a BP level and CVD risk over a fixed threshold, 2) counts these individuals (number of eligible subjects, NES), 3) simulates their exposure to BP-lowering drugs and 4) estimates the number of events prevented (NEP) among them[2]. We assumed that the effect model that describes the relationship between the risk of CVD death without any treatment and the same risk modified by BP-lowering drugs[18] is linear. For each individual on BP-lowering drugs, the risk on treatment was obtained by multiplying its baseline risk by an average 0.83 relative risk of fatal CVD for both men and women derived from a meta-analysis of BP-lowering drugs trials[19], assuming no secular trend of effects (Fig 1). Health benefits were measured as the life-years gained[20], computed as the product of the number of events prevented by the treatment[2], and the 10-year life expectancy (Eq 1).

Download:

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

Fig 1. Procedure used to determine the proportions eligible and the number of events prevented thanks to treatment.

Abbreviations: CVD, cardiovascular death; NEP, number of events prevented; NES, number of eligible subjects (Modified from reference 2).

https://doi.org/10.1371/journal.pone.0140793.g001

(1)

Eq 1 allows computing the life-years gained over 10 years of treatment exposure. The benefit from treatment is the number of events avoided multiplied by the 10-year life expectancy. NEP and LE are, respectively, the number of events prevented and the mean life expectancy in the category.

We used the number of eligible subjects over a 10-year period as a proxy for costs and computed the efficiency as the number of subjects needed to treat over ten years to gain a life-year.

Designing a new treatment strategy

We estimated the years of potential life lost—YPLL[21] from the number of cardiovascular deaths that would occur if no treatments were administered among the individuals potentially eligible and the 10-year mean life expectancy (Eq 2).

(2)

Eq 2 to calculate years of potential life lost due to premature death of cardiovascular origin. CVD: fatal cardiovascular disease; LE: mean life expectancy estimated from French official statistics reported in 2010[22]; i: the individuals in the category.

We defined the proportional benefit as the ratio of the potential gain in life years under treatment over the expected years of life lost due to premature death of cardiovascular origin (Eq 3).

(3)

This proportional benefit estimated at the population level served as the reference for the equitable strategy. We computed the proportional benefit coefficient from the 2007 ESH/ESC guidelines implementation in the individuals potentially eligible of all ages and both genders, using the Eq 4: (4)

Computing proportional benefits for each category of individuals

In order to allocate the same proportional benefit to individuals of all categories of age and gender, we multiplied the burden of years of potential life lost class-specific by the PB_ESH/ESC coefficient to estimate the desired life-years gained (Eq 5, derived from Eq 3) and the number of events to prevent (NEP_des) under the PB strategy (Eq 6, derived from Eq 1). Thus, among the youngest potentially eligible male about 400 deaths of cardiovascular origin were expected to occur in the next ten years, while 1600 deaths were expected in the second age category and 4000 deaths in the third category. Applying a PB coefficient of 10% implies that the desired number of deaths prevented would be, respectively, 40, 160 and 400.

(5)

Eq 5 to compute the desired life-years gained by categories of gender and age according to the Proportional Benefit approach. It corresponds to the years of potential life lost multiplied by the overall ESH/ESC proportional benefit coefficient.

(6)

Eq 6 used to calculate the desired numbers of events prevented over 10 years by age and gender categories, where NEP: number of events prevented; LYG: life-years gained; LE: mean life expectancy.

Identification of risk thresholds adjusted by gender and ages

The function that simulates the exposure to treatment (Fig 1) was used to find the risk percentile that allows achieving the benefit desired NEP_des within each category of individuals potentially eligible. We tested by iteration an increasing risk percentile into the function until the NEP obtained was equal to the NEP desired (Fig 2). The resulting Proportional Benefit strategy thus recognizes the individuals to treat from the new CVD risk thresholds adjusted by ages and genders. In our example, these thresholds corresponded to percentiles 65, 63 and 61 in the first, second and third age category of virtual male.

Download:

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

Fig 2. Simulation steps to design the proportional benefit strategy to identify the treatment target population.

The reference is the overall Proportional Benefit, i.e. the relative life-year gain obtained from the application of the 2007 ESH/ESC recommendation on the potentially eligible individuals from the Realistic Virtual French Population. Abbreviations: PB, proportional benefit; YPLL, years of potential life lost; LYG, life-years gained; NEP, number of events prevented; NNT, number needed to treat to gain a life-year.

https://doi.org/10.1371/journal.pone.0140793.g002

Comparing both treatment strategies and benefit ratios related to gender

We compared the PB strategy to the European recommendation of 2007 in terms of numbers of life-years gained and numbers of subjects needed to treat to gain a life-year upon ten years of treatment exposure. To account for the opportunity cost of investing at present rather than in the future considering that people give greater value to current health benefits over future benefits[13], the computations were done with and without discounting costs and benefits. A time preference rate of 3% per year was applied[14,23] to the 10-year life expectancy potentially recovered according to Eq 7. Future costs were discounted at the same yearly rate as health benefits accordingly with current NICE guidelines[12], assuming a constant cost-effectiveness threshold over time[9].

(7)

Eq 7 To compute time preference rate applied to the life expectancy after 10 years of treatment exposure. With r: time preference rate and LE: life expectancy.

As a measure of how much the PB strategy reduces the differences of benefit between genders obtained applying the ESH/ESC recommendation, we computed the men/women ratios for the years of potential life lost without treatment and the benefits expected, expressed as NEP and LYG, at each 10-year age interval.

Sensitivity analysis

The 2013 guidelines approach indicates that treatment is recommended if evidence is available and it may or should be considered when evidence is weak or unavailable. This is the case of grade 1 hypertension elderly patients and grade 1 hypertensives at low risk. We performed a sensitivity analysis to test to what extent the new recommendations could modify the proportions treated and benefits across the different categories of the potentially eligible population compared with the previous guideline version of 2007. We simulated the implementation of the new rule and assessed its impact on the NES and NEP, the number of residual events, i.e. the events that remained untreated in the individuals considered ineligible plus the events that remained despite BP lowering in treated individuals, the proportion of benefit and the efficiency. Elderly patients were identified as 60-year-olds and older individuals[24].

Ethics Statement

Not applicable.

Sensitivity analysis

Implemented on the same population of potentially eligible individuals, the scenario with the lowest level of evidence according to the 2013 ESH/ESC guidelines, i.e., the opinion-based scenario, dramatically amplified the NES among the lowest risk individuals with 1472-fold more women to treat under the age 45 years, whereas the evidence-based scenario reduced to less than a half the NES among the oldest men and women compared to the 2007 guidelines (results not shown). In individuals of both sexes younger than 55 the NES was amplified by three under the opinion-based scenario, and virtually maintained by the evidence-based scenario, compared to the 2007 ESH/ESC guidelines. The LYG under the opinion-based scenario was 1,5-fold the LYG expected under the 2007 guidelines before the age 55 years; whereas in individuals aged 55 years and older the evidence-based scenario reduced this gain to one-third the LYG expected under the 2007 guidelines. The proportion of benefit, i.e. the relative life-year gain, was augmented in individuals younger than 55 years in the opinion-based scenario and reduced by more than two thirds at ages over 55 years in the evidence-based scenario, compared to the 2007 guideline version. The overall proportion of benefit from the 2007 guidelines was halved by the 2013 guidelines evidence-based scenario and increased by the opinion-based scenario (Table 2). The residual number of events was globally increased with the evidence-based scenario and decreased with the opinion-based scenario compared with the 2007 guidelines. The number needed to treat to gain a life-year over 10 years of drug treatment was greater under the 2013 guidelines disregarding the level of evidence of the scenario adopted compared to either the 2007 guidelines or the PB strategy (Table 2).

Strengths and limitations of the study

Our strategy considers specific risk thresholds for age-sex categories as the unique additional rule besides BP levels to decide for treatment, instead of the various categories of individuals to treat proposed by the 2007 ESH/ESC guidelines. Simplified rules have the potential to improve physicians’ compliance to recommendations reducing the important gaps between recommendations for CVD management and actual clinical practice reported by observational studies[25].

As the benefit depends on the burden of disease, there is no reason to fix a unique budget for preventing CVD in every population and a re-calibration of risk estimation systems is promoted by the ESH/ESC guidelines to allow for secular trends in CVD mortality in different countries. Our approach allows adjusting the eligibility to treatment to the specific population background in terms of demography, the burden of disease and resources instead of calibrating the risk score in support of a suitable application of the absolute risk rule.

The PB strategy results in lower therapeutic coverage of people at high risk compared to the 2007 ESH/ESC strategy; of note, under this strategy, 67% of treated individuals of both genders were older than 55. However, according to the latest ESH/ESC guidelines, in elderly patients drug treatment should be initiated when SBP is greater or equal to 160 mm Hg. This implied that at the age of 60 years and over no one would have an indication of drugs among the mild hypertensive individuals included in our analysis and the proportion of 55-year-olds and older among the individuals treated decreased to only 24%. The PB strategy occupied a position in-between.

Regarding equity, concerns have been made on whether resources should be concentrated on the patients who can benefit most[26]. In a modelling study that estimated the benefits and costs of different prevention activities to reduce the burden of CVD in the US, lowering LDL cholesterol in people with existing coronary artery disease offered the largest benefit per person in terms of absolute myocardial infarction risk reduction. However, for the population as a whole this benefit was relatively small compared to other preventive activities given the reduced number of individuals who were candidates to this measure[27].

Main consequences of the equitable proportional benefit strategy

The allocation of a constant relative benefit whatever the age or sex of the individuals is in line with the evidence regarding the effect of BP lowering drugs. Two recent meta-analyses of BP lowering trials stratified by levels of cardiovascular risk [28,29] indicate that the relative reduction of cardiovascular events obtained by BP lowering does not change at different baseline levels of cardiovascular risk and therefore, the absolute benefit increases with increasing risk. The PB strategy maximises gains per categories by allocating drug treatment to the individuals at the highest risk relative to their peers.

The PB strategy increases the therapeutic coverage of people with relatively low risk, i.e. young and women, who have also the greatest life expectancy. As an obvious consequence, the gain in life-years was increased in this subgroup, while the overall number of deaths prevented was reduced compared with the 2007 ESH/ESC guidelines that concentrate prevention efforts on the oldest men (Table 2). Considering the expected age-dependent decrease in QALYs[30] given the lifespan reduction and assuming that utility values per health state decrease with age and co-morbidities, redistributing the number of deaths prevented under the PB strategy would result in greater gains of quality-adjusted life years compared to the ESH/ESC recommendation. Opposite to the results of our equitable treatment rule, Liew and colleagues[14] have suggested that for individuals with different life expectancies but the same short-term CVD risk, treatment makes little difference on health benefits due to the counterbalancing effects of time preferences, case fatality rates and competing risks. It is thus necessary to provide a definition of what benefit is. From a societal perspective, discounting life expectancies by time preference rates implies giving relatively higher value to elders’ lives compared to the remaining lives of younger individuals, as indicated by the reported 40% reduction in life expectancies after discounting at ages 35–39 (from 41.7 to 23.6 years) compared to an 8% reduction at 85 years and over (from 4.87 to 4.47 years). But does the naturally higher vulnerability to death of the individuals that have already lived longer really make worthier treating them than treating younger individuals?

Based on the 2007 ESH/ESC recommendations, the PB strategy considers initiation of antihypertensive treatment when BP is in the high normal range and CVD risk is high compared to the peers’ risk. Drug treatment of high normal BP individuals has proven to be efficacious to reduce the incidence of hypertension with promising substantial public health effect, but the evidence in favour of early interventions is still considered limited. The recommended lifestyle measures for BP control in prehypertension have no demonstrable effect on public health. Assessing how far the benefit of early drug intervention lasts is key to address the issue of treatment eligibility in high normal BP subjects.

Modelling studies have shown that treating hypertension is cost-effective in different populations[31,32], with greater efficiency at increased age of treatment onset[32]. This is not surprising since age increases cardiovascular risk and thus maximises the gains per cost unit. The efficiency, as defined in the present study, was reduced with the PB strategy compared with the 2007 European recommendation, because a greater number of individuals have to be prescribed BP-lowering drugs to extend the benefits to the youngest and especially women. Nevertheless, it has also been shown that for persons treated with hypertension interventions, life extension could be achieved without increasing average lifetime medical spending[31]. Moreover, reducing morbidity as a result of hypertension treatment would still reduce health spending due to the avoided costs of informal care after an event[32].

The PB strategy offers an intermediate point of view between the classical high-risk approach promoted by historical guidelines and the latest hypertension guidelines that introduce new evidence-based recommendations. While the classical approach is the most efficient, the forecasted effectiveness and efficiency of the newest guidelines appear to be hardly competitive against this widely accepted reality. Our original approach has the merit of redistributing the same overall theoretical benefit of the high-risk approach within a population providing the same proportional benefit to every category of individuals, while improving the profile of residual events in younger individuals without elevating costs dramatically. The redistribution of benefit that reduces the residual risk at younger ages is in line with the conclusions of Thomopoulos et al., which promote extending treatment to low-to-moderate risk hypertensive patients in order to prevent the failures due to delayed onset of treatments[28].

Implications for clinicians and policymakers

The current approach combines the absolute risk rule captured by the overall benefit expected from the application of the 2007 ESH/ESC guidelines used as a constraint to design the PB strategy with the application of the peers’ risk previously suggested. Other indices could also serve as constraints, such as the resources allocated to cardiovascular prevention at the scale of a country population, thus determining the maximum allowed number of eligible subjects to be distributed across the pertinent categories of individuals according to the proportional benefit definition.

The PB strategy is based on the discrimination ability of the risk equations used and, similarly to the risk age approach[1], calibration is indeed irrelevant to the rule application. Discrimination has been shown to be consistent across populations despite their different absolute risks[33–40] extending interestingly the potential applications of our methodology to other populations, disregarding the risk estimation system used as long as it has reasonable discriminatory power. Adapting the treatment rule needs to identify the CVD risk thresholds class-specific that lead to save a prefixed proportion of life-years in the years of potential life lost across the categories of individuals; even if the risk equation overestimates or underestimates CVD risk, this proportion would be unaffected. The CVD risk of the individual patient could be referred to the treatment rule established for that risk predictor in the whole country population until suitable coefficients are identified to calibrate the prediction.

Perspectives

Our approach has the merit to highlight several dimensions in the decision process, which can be considered as independent: 1) effectiveness, 2) economics, which is combined with the first dimension into efficiency aspects; and 3) level of evidence: this dimension is illustrated by the fact that when guidelines promoted treating people with BP between 140/90 mm Hg and 160/95 mm Hg, any good level of evidence was lacking[41]. Our modelling approach should take this dimension into account in further developments. In next steps, we envisage demonstrating the generalizability of our approach by adapting the current recommendation to identify the individuals eligible to treatment in the Chilean population.