Does Polypharmacy Lead to Nonadherence or Nonadherence to Polypharmacy?
See related article, pp 1113–1120
As Gupta et al,1 Tomaszewski and collaborators describe that patient adherence is a major problem for controlling blood pressure (BP) and is difficult to accurately assess in practice and research. One method for assessing adherence is to biochemically detect prescribed BP medications or metabolites in blood or urine samples. There is increasing interest in this method because it has already identified very high levels of nonadherence in hypertension referral centers in Europe and the United States.
The article by Gupta et al1 reports on the largest sample to date of patients tested biochemically for adherence. They found that the strongest predictor of nonadherence is the number of medications prescribed. Other predictors are being female and a younger age. On the basis of their data, the authors propose a predictive model for detecting nonadherence in settings where biochemical testing is unavailable, and they postulate that the use of fixed-dose combinations of antihypertensives, which can reduce the numbers of pills prescribed, may alleviate nonadherence. These are all noble goals and, to many, seem both intuitive and benign. However, a few caveats and limitations emphasize that further study is critical before any clinical implementation is attempted.
The generalizability of the model presented is severely limited by the sample from which it was derived. The key observation (ie, that nonadherence is associated with the number of medications) came from patients who were referred to a specialty center for suspected nonadherence, presumably because of poor BP control. What type of patients would typically be referred to this type of study? A practicing MD likely would not refer the following patients to this type of study: those who did not receive consistent, active physician care; those who did not take medicines consistently because of side effects; those whose medications did not adequately control BP; those whose BP was considered adequately controlled; or those who were suspected to have comorbid conditions. The investigators did not report data on other very important variables that should be available for all patients treated with multiple drugs and could make BP control more difficult, for example, poor renal function. More importantly, the observation that nonadherence is associated with the number of medications by itself may not lend support to the use of combination drugs. Indeed, it is likely the failure to control BP with multiple medications—medications that were sequentially added because of continued high BP despite the previous intensification of therapy—that would most likely elicit a referral.
This is not at all a trivial or strictly theoretical possibility. There is clear evidence in the literature that medical regimens are intensified in nonadherent patients when they fail to reach a goal.2,3 In reality, practitioners often do not know which of their patients are nonadherent. Indeed, if a practitioner only suspects nonadherence after a patient reaches 6 medications, that MD may be a bit too trusting in humanity.
The model had modest predictive value when applied to a second group of patients. However, these patients were from a hypertension referral center. It is important to acknowledge that for any specific disease, only a small fraction of patients reach university specialty clinics,4 and their referrals are not random. The major driver of these referrals is typically that the patient’s BP is not controlled despite the primary MD prescribing multiple drugs. Very high rates of biochemical nonadherence are found in these populations. By contrast, the adherence rate in nonreferred populations with apparently resistant hypertension is actually low and better than that of other hypertensives in the same health systems.5 However, BP in the uncontrolled patients of this population is much closer to being controlled than it is in the referral population, and adherence is measured from pharmacy refill records. Thus, the accuracy of the predictive model may be very different when applied to a nonreferred population. Incorrectly diagnosing a patient with nonadherence might inappropriately alter the treatment plan and seriously challenges the doctor–patient relationship.
The authors suggest that their data support the idea that decreasing the number of pills that a patient takes by combining therapeutic doses of ≥2 antihypertensives will help improve adherence. Indeed, many reviews and treatment guidelines recommend this strategy. Nonetheless, the authors do point out that not all studies suggest that adherence declines as the number of medications increases. Patients who take medication for hypertension likely also take medications for other conditions. Thus, it is likely that the total number of pills, not just the number of BP pills, matters to patients. A recent meta-analysis of studies using the total number of medications used for all conditions showed that adherence declined with increasing numbers of medications in only 2 of 6 studies.6
In the field of hypertension, there has been an increased interest in fixed-dose combination pills, and a large fraction of the few hypertensive drugs in the developmental pipeline are combination pills. There are many observational studies and some clinical trials7 comparing fixed-dose combination pills to single-medication pill strategies. Most of the observational studies use statistical methods to adjust for nonrandom placement of patients in these treatment strategies and have modest increased adherence, albeit often without any improvement in BP control. Most studies were conducted on new patients and not in individuals whose conditions have proven difficult to control on multiple agents. However, the overall picture from these studies is not entirely clear, with 2 of the largest and most recent observational studies showing that patients who were initially prescribed 2 single agents had better adherence than those prescribed 1 combination pill.8
In addition, nonrandomized trials risk an unidentified covariate of combination therapy confounding the results, which is particularly problematic in this field because the overwhelming majority of patients do not receive combination drugs. Therefore, those who are prescribed combination therapy might represent slightly different patient populations. The small number of randomized trials, which should eliminate the effect of unmeasured confounding variables, do not show a consistent advantage of combination pills over multiple single-medicine pills.9 Indeed, a meta-analysis of 7 randomized controlled trials involving 397 patients showed no significant difference in the achieved BP, the outcome of real interest. Only 3 studies attempted to measure adherence, and 2 of these showed no differences. The authors conclude that the low quality of available evidence does not confirm or rule out a substantive difference between fixed-drug combinations and free-drug combination therapy in the management of hypertension.
In addition, fixed-dose combination pills may not be harmless.10 Some are expensive, and they do not allow the flexibility needed for titration. If a patient stops taking the medication, both drugs may be eliminated. In addition, of the many combination medications produced, very few contain chlorthalidone and spironolactone, which are valuable in patients with difficult-to-control BP. Another potential problem with using combination pills is, as the authors show, that most nonadherence is partial. Most patients take some of their prescribed medicines. It is not clear that, in these patients, getting 3 of 5 medications is worse than getting none of 1 medication. Patients with difficult-to-control BP on multiple medications are a special challenge, and adherence is likely to be only part of the problem.
Biochemical testing for adherence is not yet widely used, and its role will likely evolve. As the authors point out, biochemical testing itself is imperfect. It detects low levels of a drug or metabolite not therapeutic levels. Thus, a drug with a long half-life could be present, even if not taken regularly. The authors acknowledge that biochemical testing likely underestimates actual nonadherence. Moreover, as patients become more familiar with testing and it loses its element of surprise, patients with complex agendas will quickly learn to take pills before a visit, further reducing the value of the testing. As the test becomes more widely used, its cost will likely decrease, although its cost is already in the range of other tests used to evaluate patients with severe hypertension from secondary causes. Thus, more research is needed to understand how to best treat this high-risk group and how to address medication nonadherence.
The opinions expressed in this article are not necessarily those of the editors or of the American Heart Association.
- © 2017 American Heart Association, Inc.
- Gupta P,
- Patel P,
- Štrauch B,
- et al
- Heisler M,
- Hogan MM,
- Hofer TP,
- Schmittdiel JA,
- Pladevall M,
- Kerr EA
- Wimmer BC,
- Cross AJ,
- Jokanovic N,
- Diug B,
- Bell JS
- Gupta AK,
- Arshad S,
- Poulter NR
- Filippone EJ,
- Foy AJ
- Mallat SG,
- Tanios BY,
- Itani HS,
- Lotfi T,
- Akl EA