“Real‐world data analysis” in disease management such as hypertension: Has the time come?

Presently, evidence‐based medicine is required to significantly influence the management of disease states such as hypertension. Specifically, evidence‐based modalities include hypothesis‐driven randomized clinical trials (RCT) (the gold standard) as well as appropriately performed meta‐analyses, population studies, and/or clinical effectiveness studies. Hypertension or an elevated blood pressure is a leading risk factor in morbidity and mortality secondary to a variety of cardiovascular diseases including of the heart (ischemic heart disease, congestive heart disease, left ventricular hypertrophy, and atrial and ventricular arrhythmias), cerebrovascular disease (stroke and dementia), renal disease (chronic and end‐stage renal failure), and peripheral vascular disease. In fact, hypertension is now recognized as the leading risk factor for cardiovascular disease worldwide and its prevalence is increasing, especially in low‐ to middle‐income countries.1, 2 Fortunately, there is both safe and effective non‐pharmacologic (lifestyle modification) and pharmacologic treatment for hypertension. Such treatment has been guided by the results of landmark RCT, such as the DASH dietary and sodium studies,3 the Veterans Administration Cooperative Studies,4, 5 the Hypertension Detection and Follow‐up Program,6 the Systolic Hypertension in the Elderly Program,7 the ALLHAT trial,8 and more recently the SPRINT trial.9 However, despite this body of evidence‐based medicine and the availability of safe and effective treatment, the control rate of hypertension (even defined at a systolic blood pressure of less than 140 mm Hg and a diastolic blood pressure of less than 90 mm Hg) is only about 50%‐70% in high‐income countries and as low as 14% globally,10 particularly in low‐ to middle‐income countries.

Clearly, there are many reasons for these rather dismal hypertension control rates, given the resources and evidence‐based medicine available. Such reasons include, but are not limited to, the lack of available and affordable antihypertensive medications or diets low in sodium, calories, fats, and refined sugars as well as high in potassium, magnesium, and fruits and vegetables; inadequate education as to the significance of hypertension; clinical or therapeutic inertia; the lack of use of a standardized formulary and simple, straightforward treatment algorithms; and limitations in clinical access. These reasons, and others not listed, point to other important and at times obvious factors. These factors include the observations that the individuals and populations studied in RCTs as well as the rigid methodology necessary to conduct well‐done trials do not represent the real‐world setting of treating individuals with hypertension. A vivid example of such a difference is seen in drug safety and efficacy. Since the RCTs needed for formal drug approval from agencies such as the Food and Drug Administration in the United States are indeed limited and consist of a small number of selected patients, post‐marketing surveillance of an approved drug in the real‐world setting is essential to determine further safety and effectiveness of an approved medication. There are numerous examples where the real‐world data differed significantly from the RCT data and interestingly in both directions; that is, the real‐world data demonstrated a new safety issue or a new benefit. For example, in 2004, rofecoxib was withdrawn from the market when post‐marketing data suggested a possible association with an increased risk of myocardial infarction and stroke.11 In the other direction, several new glucose‐lowering agents have shown unexpected benefit including the glucagon‐like peptide‐1 receptor activators12 and the sodium‐glucose cotransporter 2 inhibitors13 both of which have demonstrated a reduction of cardiovascular disease end points in post‐marketing surveillance. In both of these cases, the real‐world data did not match the trial data.

Thus, we may be at a critical time where the increasingly sophisticated electronic medical records (EMRs) that encompass larger patient populations, coupled with mergers creating larger integrated health systems, and expanding health maintenance organizations and clinic system models of health care such as the Mayo Clinic, Cleveland Clinic, and the former Scott & White health care systems, are available for innovative data analysis and clinical use. Hypertension is a disease particularly suited for analysis using real‐world data. Hypertension is prevalent worldwide, has an easy‐to‐identify and measure phenotype, is recorded almost universally in medical records, and has a high associated morbidity and mortality. Thus, hypertension seems an ideal target for the use of real‐world data analysis.

Such a new and innovative analysis of real‐world data in the treatment of hypertension is exemplified in the article authored by Stapff and Hilderbrand14 entitled “First line treatment of essential hypertension: A real world analysis across four antihypertensive treatment classes” and published in this issue of our journal. The authors used data available from the EMR in the TriNetX system to compare patients starting and adhering to antihypertensive treatment with four different classes of antihypertensive agents: diuretics, beta‐blockers, renin‐angiotensin inhibitors (angiotensin‐converting enzyme inhibitors or angiotensin receptor blockers), or calcium channel blockers. All patients studied had to be on the single agent for at least three years. Patient demographics were recorded, as was blood pressure and heart rate, various laboratory parameters, and cardio‐ and cerebrovascular events/outcomes. TriNetX is a global federated research network providing access to statistics on EMR from approximately 60 million patients in 50 large health care organizations predominantly in the United States. An advantage of the TriNetX data repository is that it represents a wide range of geographies, ages, and income levels as well as a mix of hospital, primary care, and specialty treatment providers.

Within the TriNetX system, there were almost six million patients with the diagnosis of hypertension with approximately three million having a cardiovascular medication started after 12/31/2008. Of those, 204 865 patients had documentation of one of the four antihypertensive treatment classes of three or more years, with 79 288 (30%) of those fulfilling the requirements for stability within one of the four classes of antihypertensive agents for at least three years. Interestingly, within those, 17.4% were on diuretics, 25.9% on beta‐blockers, 45.1% on renin‐angiotensin system (RAS) inhibitors, and 11.6% on calcium channel blockers.

The results of the analysis indicated that black patients were more likely to be on diuretics or calcium channel blockers; females were more likely to be on diuretics; patients with ischemic heart disease were more like to be on beta‐blockers; those with diabetes were more likely to be on a RAS inhibitor. Over the course of three years, patients taking RAS inhibitors had the fewest occurrences of a blood pressure above 140/90 mm Hg (28.7%) and beta‐blockers the most (45.9%). RAS inhibitors lowered blood pressure the most—on average by 5.8/3.8 mm Hg and beta‐blockers the least—by 3.6/2.2 mm Hg. Diuretics and calcium channel blockers fell in the middle, lowering blood pressure by 4.9/3.1 and 4.8/3.2, respectively. Those patients taking RAS inhibitors or diuretics experienced the lowest rates of the predefined clinical outcome events (first stroke, any cerebrovascular outcome, myocardial infarction, any ischemic heart disease, chronic kidney disease). Beta‐blockers were also associated with the highest rates of cardiovascular events, mostly myocardial infarction and ischemic heart disease. The authors comment that the beta‐blockers’ higher cardiovascular event rate could partially be due to a greater number of patients with preexisting heart disease preferentially receiving beta‐blockers. However, even after adjusting for preexisting cardiovascular conditions using propensity score matching, patients taking beta‐blockers still had a greater number of cardiovascular events compared to those in the other three classes. Calcium channel blockers showed the highest rate of cerebrovascular outcomes; beta‐blockers closely followed.

While the authors point out that the actual outcomes reflected in the TriNetX data should not be used to compare the four classes as a RCT would, it is reassuring that the real‐world prescribing patterns correlate with most current hypertension treatment guidelines. For example, blacks were more likely to be on diuretics and calcium channel blockers as opposed to RAS inhibitors and beta‐blockers. These real‐world data match The International Society of Hypertension in Blacks’ consensus statement that certain classes of medications, specifically diuretics and calcium channel blockers, lower blood pressure on average more than beta‐blockers and RAS inhibitors in blacks when used alone.15 Females were less likely to take RAS inhibitors, possibly, as the authors noted, due to teratogenic issues surrounding conception. Individuals with diabetes were more likely to be on RAS inhibitors; the evidence‐based recommendation for hypertension in patients with diabetes and proteinuria is a RAS inhibitor.16 Other data are more circumspect such as the observation that individuals taking calcium channel blockers had higher baseline blood pressures than those on the other three classes. This could possibly explain the higher incidence of cerebrovascular complications such as stroke seen in the calcium channel blocker group, since strokes are closely linked with an elevated blood pressure. Neither the Eighth Joint National Committee17 nor the American College of Cardiology/American Heart Association guidelines18 recommend beta‐blockers as first‐line antihypertensive therapy since they may not decrease some cardiovascular outcomes as much as the other three classes. Interestingly, the real‐world data in this paper showed that those on beta‐blockers were more likely than the other three classes to have an adverse cardiovascular outcome, supporting current guidelines.

Because many of the results of this analysis seem to confirm what years of RCTs have shown, it is time to consider using large repositories of data to a greater extent in the future to determine adherence to current guidelines as well as to guide clinical interventions and decisions. Using real‐world data—not just from EMRs but also from smartphones, activity trackers, health surveys, insurance claims—could ultimately save money and increase efficiency. It is important to implement EMRs that are easy to use but also contain real‐world data on multiple factors (ethnicity, weight, family history, adherence, among others). If we move forward with this direction, it is also important to recognize the limitations of using large data sets, such as the lack of individual randomization and the limitations of the use of ICD codes for disease documentation including comorbidities. Also, drug dosages are difficult to determine. Given a balanced understanding of these issues, the authors demonstrate the distinct ability and utility of using real‐world data to complement the evidence‐based medicine gathered in rigorous RCTs and show that studies using EMR data can complement RCTs. It is hoped that the publication of such studies will spark an active discussion surrounding this issue as data get more robust and better delineated. We welcome that discussion and believe the time has come to move forward.

CONFLICT OF INTEREST

The authors have no conflicts of interest to declare.