The Emperor Has No Clothes: Dementia Treatment on the Eve of the Aging Era

A recent article in the New York Times has drawn attention to the wide range of expert opinion regarding the efficacy of acetylcholinesterase inhibitors (AChEIs).1 The opinions quoted ranged from "overwhelming evidence" to frank skepticism. This apparent lack of consensus flies in the face of U.S. Federal Drug Administration (FDA) approval of multiple AChEIs and memantine (a noncholinergic second-generation medication); multiple evidence-based reviews confirming the efficacy of AChEIs; the promulgation of official guidelines endorsing the use of AChEIs for the treatment of dementia (e.g.,2); the incorporation of such guidelines into endless local treatment algorithm, the threat of malpractice suits and "other monetary penalties" attendant to the failure to use AChEIs, based on its perception as a standard of care;3 and suggestions that the use of placebo arms in Alzheimer's disease (AD) clinical trails is now an unethical practice.4 How can clinical skepticism exist in the face of such overwhelming institutional support for AChEIs? Well, you get the answer to the question you ask. The evidence is overwhelming that AChEIs have statistically significant effects on certain clinical outcome measures. Whether these effects are also clinically relevant, and whether they justify the associated costs, require value judgments that cannot be so easily quantified. Certain points are important to remember when approaching this literature. First, statistical significance is a reflection of statistical power. The drug's effect size, measurement error, and sample size, influence power. Clinical trial sponsors cannot easily modify their drug's effect size, but sample size and measurement error can be manipulated. Given enough resources, large samples can show small effects to be statistically significant. A corollary to this is that the larger the sample size, the smaller the effects that can be detected. When reading large studies with thousands of participants (the Women's Health Initiative comes to mind), one should consider that the statistically significant effects may be clinically trivial or exceedingly rare. In addition to the sample size, trial designers also have control over the outcome measures. Some have been specifically engineered for their sensitivity to change, often at the expense of clinical utility. Thus, measures such as the Alzheimer's Disease Assessment Scale Cognitive subscale (ADAS-cog),5 which are frequently used in clinical trial work, are rarely employed outside of that context. Clinicians are less familiar with these instruments and may not be comfortable interpreting their scores. For example, a delay in initiating AChEIs has been widely reported to significantly limit gains in cognition relative to subjects who are treated from study inception,6 but this conclusion is based on studies that used the ADAS-cog as an outcome measure.7,8 Subjects who delayed AChEI exposure by participation in the placebo arms of earlier protocols appear less responsive than those treated continuously at full dose, but the differences across such groups is only on the order of about 1.5 ADAS-cog points. Each point on that measure has been estimated to represent only 0.5 points on the more familiar Mini-Mental State Examination (MMSE).9 Another source of confusion is the very real limitations on a healthcare provider's ability to accurately diagnose AD. Lewy body dementia (LBD), for example, is an AD-like dementia estimated to account for 20% of dementia in the elderly population. Even experts may misdiagnose LBD as AD.10 LBD also appears to respond better to AChEIs than AD.11 Thus, LBD may disproportionately account for the dramatic responses encountered from time to time when treating "AD." Similarly, 40% to 60% of clinically diagnosed "AD" cases have comorbid ischemic cerebrovascular disease (ICVD).12 Clinically "silent" frontal system lesions represent the majority of ICVD detected using magnetic resonance imaging.13,14 Such lesions contribute to disability independently of certain cognitive measures, and are inversely related to the severity of AD pathology in demented cases.15 This suggests that the "AD" diagnosed by clinicians is often mixed at best. In fact, and in contrast to widely cited statistics, the clinical diagnosis of AD is uncertain in the presence of ICVD16 or after the age of 79 (i.e., the most prevalent segment of dementia cases and one at a disproportionate risk of comorbid ICVD).17 Nevertheless, subcortical ICVD has recently been shown to disrupt cortical cholinergic markers, even in the absence of comorbid AD pathology.18 Thus, the AChEI response of mixed cases may disproportionately reflect the contribution of their ICVD.19 Ironically, diagnostic uncertainty suggests the possibility that the AChEIs, indication for "AD" may be founded on the presence of comorbid non-AD pathology in a subset of cases. This should temper confidence that the results of evidence-based reviews are specific to "AD," or that strict adherence to FDA-approved indications is clinically meaningful. The search for effective dementia treatments is likely to become even more pressing as the baby boomers begin their retirement. Even small reductions in the average age of onset of AD are capable of greatly reducing the projected exponential rise in dementia cases.20 We are going to need all the data at our disposal to improve the clinical utility of our prescription practices. I believe the "evidence based" movement, which promulgates guidelines for evaluating the quality of the literature supporting a particular drug, is stifling this search. These guidelines place a high value on statistically significant effects in double-blind, placebo-controlled trials and yet generally fail to consider the more subjective and value-driven clinical utility of these effects. Thus, despite the "overwhelming evidence" in support of AChEIs, there is no consensus among clinicians regarding when to start AChEIs or, more significantly, when to stop. Similarly, the claim that AChEIs delay nursing home placement is based on uncontrolled analyses of questionable validity.21–23 Memantine's role is even less clear. Clearly, the relative immaturity of this literature can hinder evidence-based approaches, but it is unrealistic to suggest awaiting the outcomes of future studies. Such studies may never be done. Tacrine was approved in 1993, yet there is still not a single placebo-controlled head-to-head comparison between two AChEI. Meanwhile, the use of second-generation dementia treatments (e.g., memantine) is being restricted, notably by the Department of Veterans Affairs (VA).24 The VA limits AChEI and memantine to their FDA-approved indications. This effectively narrows their use to particular stages of a particular diagnosis (i.e., AD) despite the fact that "staging" of clinically diagnosed AD is crude at best; the VA's interpretation of the FDA indication as being a function of MMSE scores is itself untested by evidence-based review; it is questionable whether pivotal AChEI clinical trials actually succeeded in selecting pure AD groups;25 and the target population of clinically diagnosed veterans with "AD" is surely neither a homogeneous group nor even one necessarily dominated by AD pathology. Two solutions suggest themselves. First, interindividual differences in clinical practice should be actively encouraged, not discouraged. After all, clinician practice essentially represents a nonlinear search strategy similar to the "genetic algorithms" used in drug development and other industrial applications. Clinicians will eventually reach a consensus regarding the proper role of AChEI in dementia therapy. Their solution will reflect millions of patient-drug encounters in naturalistic settings. Second, expert bodies, such as the American Geriatrics Society, need to take the difficult step of promulgating standards for clinically meaningful effects in terms of cognition, mood, functional status, behavior, and dementia progression. Then industry and regulators can be challenged to meet those standards during drug development. I believe that few of the currently available studies would meet such standards. That outcome might make a much more powerful statement than the current evidence-based literature does. If these steps are not taken, we risk losing the respect of our colleagues and our patients. They will judge for themselves the efficacy of these treatments.