Generalizability of Blood Pressure Lowering Trials to Older Patients: Cross‐Sectional Analysis

Randomized controlled trials are used to inform clinical guidelines on the management of hypertension in older adults, but it is unclear to what extent these trials represent the general population attending routine clinical practice. This study aimed to define the proportion and characteristics of patients eligible for hypertension trials conducted in older people.

trials indicated that larger blood pressure reductions and multiple antihypertensive prescriptions may also be harmful. 5,6 Observational studies also suggested that blood pressure lowering in older patients may be harmful, particularly in patients with frailty and multimorbidity, increasing the risk of falls, 7 acute kidney injury, 8 and death. 9,10 The two largest trials that have examined the efficacy of antihypertensive treatment in patients aged 80 and older are the Hypertension in the Very Elderly Trial (HYVET) 3 and the Systolic Blood Pressure Intervention Trial (SPRINT). 4 HYVET showed in patients with a baseline systolic blood pressure of 160 mm Hg and higher that blood pressure lowering to a target of 150 mm Hg systolic reduces the risk of CVD, heart failure, and death from any cause. 3 The SPRINT trial showed that in patients with a baseline systolic blood pressure of 130 mm Hg and higher, systolic blood pressure lowering to 120 mm Hg (defined using automated and mostly unaccompanied measurement methods) reduced the risk of CVD and death, 4 and this effect was sustained in a subgroup of participants aged 80 and older, although not in those with reduced cognitive function. 11 However, it is unclear to what extent these results apply to older patients in routine practice, 12,13 particularly with population-wide observational data suggesting treatment may be harmful. [7][8][9] Indeed, patients in the HYVET trial are thought to have had fewer comorbidities, 14 whereas it is unclear to what extent the levels of frailty observed in the SPRINT population 15 compared with those in the general populations of Europe and North America. 16,17 The 2018 Optimizing Treatment for Mild Systolic Hypertension in the Elderly (OPTiMISE) trial examined the safety of reducing blood pressure medications in older patients with controlled systolic blood pressure (<150 mm Hg). 18 The rationale for OPTiMISE was that patients with frailty and multimorbidity are not well represented in previous blood pressure lowering trials and may in fact benefit from fewer treatments, due to a reduction in adverse events from polypharmacy. 19,20 The present study aimed to examine this rationale by (1) determining the proportion of patients registered at a primary care practice who would be eligible for each of the HYVET, SPRINT, and OPTiMISE trials; (2) establishing what patient characteristics are associated with being eligible, and (3) comparing the characteristics of patients in the participating practices who would have been eligible for the trials with those actually enrolled in each trial.

METHODS
Detailed methods are given in the Supplementary Methods S1 (Supplementary Appendix).

Study Design
This study used a cross-sectional design, utilizing anonymized data extracted from the medical records of patients registered at general practices enrolled in the OPTiMISE trial. 18

Population
Eligible patient data were those from individuals aged 80 and older and registered at English general practices hosting the OPTiMISE trial. General practices had to be using the EMIS electronic health record system and agree to anonymized patient data being extracted for the study.

Study Outcomes
The primary outcome for the study was the proportion of patients in the participating practices who would have been eligible for the HYVET, SPRINT, and OPTiMISE trials. These trials were chosen because they represent the largest trials conducted to date examining hypertension management strategies (both prescribing and deprescribing) in older adults. Secondary outcomes were to describe the characteristics of eligible patients, to determine the proportion of individuals excluded by each eligibility criteria, and to identify independent predictors of eligibility. The characteristics of those eligible in the participating practices were compared with those enrolled in each trial. Patients were classified as eligible if they fulfilled the published inclusion and exclusion criteria for each trial (Table 1). Due to the use of data from electronic health records, it was not possible to apply some eligibility criteria to the sample population (Supplementary Table S1).

Covariates
Data relating to baseline patient characteristics (age, sex, smoking status, and body mass index [BMI]), blood pressure, cholesterol, comorbidities, and details of all prescribed cardiovascular medications were extracted. A frailty index (FI) was developed retrospectively from 26 available deficits. Frailty was categorized using the thresholds reported in the SPRINT trial: fit (FI <.1), less fit (FI .1-.21), and frail (FI >.21). Cardiovascular polypharmacy was defined as being prescribed three or more medications for cardiovascular conditions (eg, antihypertensive, statin, antiplatelet, or diabetic medications).
To examine predictors of eligibility for each trial, missing data for blood pressure and BMI were assumed missing at random and imputed using multiple imputation with chained equations. 21 Data on the characteristics of patients actually enrolled in each trial were extracted from previous publications (HYVET, SPRINT) 3,11 or obtained from the original trial data set (OPTiMISE). 18

Main Analyses
Descriptive statistics were used to define the primary outcome, the characteristics of eligible patients, and the proportion of patients excluded by each eligibility criteria. To better understand the phenotype of patients eligible for each trial, a logistic regression model was constructed to examine predictors of eligibility for each trial, with general practice (site) included as a random effect. Candidate predictors for each model were prespecified as age, sex, blood pressure, BMI, frailty, number of comorbidities, number of cardiovascular medications, history of hypertension, chronic kidney disease, diabetes mellitus, type II, myocardial infarction, stroke (or transient ischemic attack), and heart failure or coronary artery bypass graft. Comparisons between the characteristics of eligible patients and those included in the original trials were made using independent samples t tests and two sample tests of proportions. Bonferroni correction was used to account for multiple comparisons.

Sensitivity Analyses
The definition of clinic blood pressure (most recently recorded reading) was examined in sensitivity analyses where the blood pressure eligibility criteria for each trial were defined according to the mean of the last three readings documented in the medical records. Further analyses examined the proportion of patients with hypertension (defined by a coded diagnosis or prescription of antihypertensive therapy) who would have been eligible for each trial. For logistic regression models examining predictors of eligibility, sensitivity analyses were undertaken based on complete cases only, to establish the impact of using multiple imputations to deal with missing data. Further sensitivity analyses were undertaken on complete cases using a standard logistic regression model with backward stepwise selection of candidate predictors, with the significance level for inclusion set to P < .05. All analyses were conducted using STATA v.14.1 (Special Edition, StataCorp, College Station, TX).

Primary Results
Overall, 24 of 69 general practices participating in the OPTi-MISE trial were using EMIS Health software and agreed to provide data for this study. A total of 15,376 patients aged 80 and older were registered at these practices.  Figure 1). No patients were eligible for both HYVET and OPTiMISE trials due to mutually exclusive blood pressure eligibility criteria, but there was some overlap in eligibility for HYVET and SPRINT, and SPRINT and OPTiMISE ( Figure 1). Having normal or controlled systolic blood pressure was the most common reason for exclusion from each trial, but up to one-third of patients with high or uncontrolled blood pressure were also ineligible (HYVET = 30.7%; SPRINT = 5.6%; OPTiMISE = 27.2%). Diagnosis of dementia (12.6%), diabetes mellitus, type II (16.5% [SPRINT only]), or stroke (14.6% [SPRINT only]) were other common reasons for exclusion ( Figure 2). Sensitivity analyses examining blood pressure defined as the mean of the last three recorded readings identified similar proportions of patients eligible for each trial (data available from authors). Similar proportions of patients were eligible for each trial when the population was limited to those with a previous diagnosis of hypertension, with the exception of OPTiMISE that had more eligible hypertensive patients (35.0%; 95% CI = 34.1-35.9%).

Characteristics of Eligible Patients
Patients eligible for each trial were similar in age and BMI, but they had different levels of mean systolic blood pressure ( 3,11 or obtained from the original trial data set (OPTiMISE). 18 Comparisons were made using independent samples t tests and two sample tests of proportions. Standard deviations required for this analysis are taken from the variance reported in the intervention arms of the original HYVET and SPRINT trials, giving conservative estimates of the difference between groups. Bonferoni correction for significance level was P < .002. morbidities varied across trials, and the mean number of cardiovascular medications was highest in patients eligible for the OPTiMISE trial (Table 2 and Supplementary Table S3).  Figure S1). Analyses using backward stepwise selection identified systolic blood pressure and reduced frailty as factors predicting eligibility across all three trials (Supplementary Figure S2).

Eligible Patients Compared with Recruited Patients
Significant differences were found between those deemed eligible in the present study and those recruited to each individual trial (Supplementary Table S3). Patients enrolled in the HYVET trial were on average younger, with higher blood pressure, lower BMI, and fewer cardiovascular morbidities compared with the eligible population in the practices considered (P < .002). Patients in the SPRINT trial were also younger, had lower diastolic blood pressure but higher systolic blood pressure, BMI, and more cardiovascular morbidities and treatment (P < .001). Participants in the OPTiMISE trial had lower diastolic blood pressure, and fewer had a history of stroke, CVD, or prescriptions for statin and antiplatelet therapy than the total population that would have been eligible (P < .001; Supplementary Table S3).

Summary of Main Findings
This study examined the characteristics of 15,376 patients aged 80 and older registered in 24 general practices in England. We found that most patients would not have been eligible for randomized controlled trials that inform clinical guidelines 22-25 on the management of hypertension. Most of the ineligibility related to the differing blood pressure eligibility criteria used by trials attempting to answer different primary questions, with up to a one-third of individuals with high or uncontrolled blood pressure excluded. This is perhaps not surprising. However, a possible unintended consequence of these criteria was that eligible patients were less likely to be frail and more likely to be prescribed multiple cardiovascular medications or have multimorbidity. Therefore, for many older patients, no randomized controlled trials exist to directly inform treatment decisions in routine practice.

Strengths and Weaknesses
This study examined data from a large sample of more than 15,000 patients registered to general practices across the south-central region of England. Individuals were representative of the region and country in terms of age but may not be reflective of practices in other parts of the country that serve populations differing in ethnicity and social deprivation. 26 It is also possible that some selection bias may have been present because only 24 of 69 practices approached were able to participate. The eligibility criteria that accounted for the largest number of exclusions were those related to having high or low systolic blood pressure. Our definition of systolic blood pressure, based on the most recently recorded value, almost certainly differs from the standardized readings taken in the original trials. 27 However, readings reflected routine practice, and the same definition was used for each trial, and so comparisons between trials are still likely to be valid.
Our analyses were unable to consider the proportion of eligible individuals who would likely give informed consent to participate in each trial or the proportion deemed in clinical equipoise by the treating physician. Indeed, in the OPTiMISE trial, as few as 9% of those deemed potentially eligible on the basis of information held in electronic health records gave informed consent and were randomized, suggesting that these estimates are likely to be conservative. Data presented in Table 2 suggest the decision to participate may have been influenced by patient characteristics such as sex, frailty, multimorbidity, and treatment prescription. It was not possible to examine the association between ethnicity and trial eligibility due to large amounts of missing data in the electronic health records for this patient characteristic.
Derivation of the FI was post hoc, and therefore some of the deficits described in previous models were not extracted from individual electronic health records. 28 Of the 26 deficits available, each of the five domains that make up a valid cumulative deficit frailty model were covered: signs, symptoms, disease, disability, and abnormal laboratory results. 29 We therefore consider our frailty model to be valid. Models used to define frailty in each of the previous trials examined here were also derived post hoc and included a wide variety of deficits (Supplementary Table S2). 15,28,30 The current analysis offered a unique opportunity to estimate frailty and compare across trial-eligible populations using a consistent definition.

Comparison with Previous Literature
A study from the Netherlands found that in a population of patients attending a geriatric day clinic, less than one-half would have been eligible for the HYVET trial, and those eligible would have had significantly fewer comorbidities. 14 The present study found of those older patients attending routine primary care, fewer than 2% fulfilled the eligibility criteria for HYVET, although those that were eligible had comparable multimorbidity. Furthermore, our finding that 34.4% of older patients would have been eligible for the SPRINT trial was very similar to that of Bress et al., 31 who examined patients from the National Health and Nutrition Examination Survey database and found 34.6% of patients aged 75 and older would have been eligible for the trial.
The present study found that eligible populations differed significantly from those who participated in the original trials. This accentuated the differences between eligible patients and the general population. Participating patients had higher blood pressure and BMI but fewer comorbidities.

Implications for Clinical Practices
Clinical guidelines for the management of hypertension in older patients continue to recommend lower thresholds for treatment initiation and targets for patients already on medication. 22,24,25 Most suggest exercising clinical judgment in patients with multimorbidity and frailty, 23,24 and some go as far as to explicitly cite the exclusion criteria from the SPRINT trial in reference to patients where such a strategy may not be appropriate. 25 The present data show the extent to which clinical judgment may be required with as many as two-thirds of patients aged older than 80 not meeting the eligibility criteria for previous trials.
Further work is needed to better understand if differences between recruited trial participants and the general population are important in determining the safety and efficacy of treatments. Ultimately, it is not feasible to conduct randomized controlled trials in all populations, but attempts should be made to better capture patients from underrepresented groups (such as those with frailty and multimorbidity) in future trials. One approach might be to consider the use of specific inclusion/exclusion criteria designed to maximize the number of patients with frailty and multimorbidity eligible.
In conclusion, the present study found that most patients aged older than 80 would not have been eligible for large randomized controlled trials examining hypertension management strategies in older adults, albeit mainly due to differing blood pressure control thresholds. Therefore, for many older patients, no randomized controlled trials exist that directly inform treatment decisions in routine practice, and so caution should be exercised when using existing data in older patients with frailty or multimorbidity. reference 335). James Sheppard and Richard J. McManus were funded by an NIHR professorship (NIHR-RP-R2-12-015). James Sheppard now receives funding from the Wellcome Trust/Royal Society via a Sir Henry Dale Fellowship (reference 211182/Z/18/Z) and an NIHR Oxford Biomedical Research Centre (BRC) Senior Fellowship. Richard J. McManus, Jonathan Mant, Carl Heneghan, and Gary A. Ford are supported by NIHR Senior Investigator awards. Carl Heneghan also receives support from the NIHR SPCR and NIHR Oxford BRC. F.D. Richard Hobbs acknowledges part support from the NIHR SPCR, the NIHR CLAHRC Oxford, and the NIHR Oxford BRC. Oliver Todd is funded by a Dunhill Medical Trust doctoral research fellowship (reference RTF107/0117). Rupert Payne receives funding from the NIHR for polypharmacy and medicines optimization research. The views expressed are those of the author(s) and not necessarily those of the NIHR or the Department of Health and Social Care.