The term “polypharmacy” dates from before the mid-1800s. A treatise of 1857 described it as “that complex, cumbrous, unwieldy system of mixing all sorts of drugs”, and even more elegantly as “complicated drugging” 1. Since that time, polypharmacy has received general censure, and various efforts have sought to combat it. These have, however, lacked a common name in English until the coining of “deprescribing” in 2003 2. This word appeared little in print until around 2012, after which its use has increased consistently along with research about it, to the point that it has allowed bibliometric analysis 3 and structured reviews 4, 5. Not everyone has backed the movement. An insightful commentator caricatured it as “fashion accessory or fig leaf” 6 and professed that “deprescribing makes me weep” 7. Carrión-Madroñal et al.'s research 8 offers an opportunity to reflect on what more than a decade of research about deprescribing has shown. Carrión-Madroñal et al.'s cohort study of an intervention to reduce potentially inappropriate medications in Spain, LESS-CHRON, adds to the literature suggesting that structured programs can reduce the number of medications prescribed to older adults. Among 229 outpatients and 231 institutionalized patients, a multidisciplinary team applying the LESS-CHRON deprescribing tool identified 960 potentially inappropriate medications, of which 542 were stopped. (As context, this represented about 5600 starting medications, so about one in ten was stopped). Of the 542, 70 medications were reintroduced by 3 months. The authors interpret the intervention as successful, insofar as it did stop the vast majority of medications targeted for discontinuation. But it merits reconsidering the context and outcomes of deprescribing and reflecting on the future of the field. I will note a few outstanding and unresolved challenges. The subjects in the study started with an average of about 12 chronic medications. Medication lists of this length have become so commonplace that the number does not generate shock, but it does defy reason. Do older adults really require a dozen medications in order to remain healthy? How have we come to accept this degree of prescription as a norm? How could anyone ascertain the differential and combined effects of so many pills? Deprescribing efforts, by chipping away at a small part of the list, skirt these larger questions. Off the cuff, one might propose that a deprescribing program which removed half or more of the average dozen prescribed medications would yield a major effect (although six medications is still a lot). LESS-CHRON and other similar interventions have, in relatively controlled conditions, stopped about one medication per patient. Perhaps that one pill was causing many problems, but 11 remain. Beyond issues of statistical significance, you have to wonder how much effort is worth expending to “clean up” a bit of the drug list while still leaving it quite “complex, cumbrous, and unwieldy” 1. Yet it is possible that the intervention had more effects than the numerical outcome suggests. More than half (56%) of the proposed deprescribing actions were accepted, a promising process outcome that encourages maximizing the algorithms and communication with providers. The study did not consider dose reduction, which may have generated patient-level benefits that are difficult to quantify. Using the count of medications as the gold standard for safe prescribing avoids the issue of primary significance, namely the benefits and risks of the regimen for what matters for each patient. The concept of the potentially inappropriate prescription (PIM) can help target each medication for reassessment, but it presents practical challenges. Unless the medication both offers no benefit (including a placebo effect) and produces definite harm, the doctor and the patient must weigh the positive and negative consequences of the pill, taking into account what matters for the patient. Because benefits and harms entail categorically different entities (e.g., symptom reduction, prevention of future harm, side effects causing unpleasant experiences of different types, interference with the potential benefits of other medications, the hassle and expense of medications), they cannot be compared simply or objectively in economic terms. (For instance, how much myalgia is worth tolerating in order to reduce cholesterol by a certain number of points? How much would one pay to have 4/10 instead of 7/10 knee pain?) This sort of moral reasoning becomes messy quickly, and no single algorithm applies. Practically, the best test of appropriateness would require the patient to have a full knowledge of their subjective and objective condition now and in the future, with or without the medication, and to make their own decision. If there were only a single medication, a provider might conduct an n-of-one trial to ascertain these points, but even then they would need to impute future benefits and harms for chronic conditions. This persistent uncertainty produces a logical conundrum, and may explain why research about deprescribing has taken various approaches to targeting culprit medications, without consensus about what makes each one, or the total sum of prescriptions, appropriate or inappropriate for any patient. The standard for establishing the effect of medications is the randomized controlled trial. By extension, it would make sense to study deprescribing by randomizing patients to continue or to stop a medication, and blinding both patients and providers. Carrión-Madroñal et al. characterize this as ethically questionable. Perhaps so, but without a design of this type, it will remain impossible to ascertain objectively the effects of stopping medications. Those who seek to develop a science of deprescribing will continue to bump against this dilemma. (There are major practical issues as well. I attempted to conduct a double-blind, placebo-controlled discontinuation trial, and found it remarkably more difficult than I ever imagined 9). Although the authors do not state as much, I assume that one key ethical challenge in a blinded deprescribing study is the diminution of provider autonomy. In the LESS-CHRON intervention by far the most common reason (70%) for rejecting recommended discontinuation was the provider's clinical judgment. Bluntly, the person prescribing the medication felt—or knew—that they knew better than the tool. How do we interpret this? Because the study did not delineate the clinician's reasoning in each case, we can only guess. Perhaps they had already tried stopping the pill, with demonstrated poor effects; perhaps the tool ignored certain benefits and harms that mattered to the patient; or perhaps the provider wanted to follow guidelines or other wisdom that the tool did not appreciate. Perhaps—which the researchers in LESS-CHRON acknowledge—the tool was simply inappropriate for that patient, and must remain a suggestion rather than a rule. But as these tools improve, could or should they supplant human clinical judgment? In the future, research ethics will likely sanction autonomous artificial-intelligence prescribing and deprescribing decisions that have an adequate scientific basis. At that point, human subjects will become little more than consented laboratory specimens, and their providers will have no meaningful role besides functioning as “automated pill dispensers” to satisfy laws. Through this process, machine learning systems might select the “best” drugs for us human beings in terms of some quantifiable outcomes that are designated as important by their own standards, but we will lose our capacity to decide what matters to us and to defy guidelines when we see fit. When deprescribing gained popularity, I was concerned for several years that the commercial forces which survive on and profit from selling medications might perceive the movement as a threat, and thus seek to discredit researchers and clinicians who sought systematically to reduce medication burden. I was probably both paranoid and overly optimistic, or maybe the response has happened in a stealthy manner, because—despite numerous initiatives and almost 2000 publications about deprescribing—polypharmacy seems alive and well. The 2003 paper that first promoted the word 2 promised “better health outcomes for older people through reducing medications”. If only the “antidote” to “complicated drugging” were so straightforward. The author has nothing to report. The author declares no conflicts of interest. This publication is linked to a related research article by Carrión-Madroñal et al. To view this article, visit https://doi.org/10.1111/jgs.70269.
Stephen Thielke (Tue,) studied this question.