Two New Sets of Diagnostic Criteria—Which Is Right for Your Clinic?
On 19 April 2011, the National Institute on Aging (NIA) and the Alzheimer’s Association (AA) published new guidelines for the diagnosis of Alzheimer’s disease (AD) and its pre-dementia and preclinical phases (see ARF related news story on McKhann et al., 2011; Albert et al., 2011; and Sperling et al., 2011). The papers followed a previous release of draft guidelines at the July 2010 ICAD conference in Honolulu, Hawaii (see ARF conference stories). The criteria coexist with those developed in 2007 (Dubois et al., 2007) and further defined in 2010 (Dubois et al., 2010) by a separate international group of experts. How does each new set differ from the old 1984 guidelines and where do they differ from each other? What will be the immediate and long-term impact? How will they be put into practice? Both sets of criteria were discussed at the 9th Annual Mild Cognitive Impairment Symposium, held 29-30 April 2011 in Miami Beach.
On 9 June 2011, Reisa Sperling, Brigham and Women’s Hospital, Boston Massachusetts, and Marilyn Albert, Johns Hopkins University, Baltimore, Maryland, presented the new criteria for preclinical AD and mild cognitive impairment due to AD, respectively. Pieter Jelle Visser, University of Maastricht, and VU University Medical Centre, Amsterdam, The Netherlands, presented those published in 2007 and 2010 by a separate International Working Group. Stephen Salloway, Brown University, Providence, Rhode Island, gave an overview of how the new criteria are beginning to be used. Howard Feldman, University of British Columbia in Vancouver, and Bristol-Myers Squibb, Wallingford, Connecticut; and Philip Scheltens, VU University Medical Center, Amsterdam, The Netherlands, joined them for a panel discussion.
by Tom Fagan and Gabrielle Strobel
Since 1984, both clinicians and researchers have, for the most part, followed consensus AD diagnostic criteria established by the National Institute of Neurological and Communicative Disorders and Stroke (NINCDS) and the Alzheimer’s Disease and Related Disorders Association—now the Alzheimer’s Association (McKhann et al., 1984). In the intervening years, research has yielded a richer understanding of the pathological and clinical progression that leads to Alzheimer’s dementia. Many in the field wondered when diagnostic criteria would be updated to reflect the field’s new understanding that the biology of AD begins to unfold a decade or more before patients reach a stage where the 1984 criteria apply.
In 2007, the International Working Group for New Research Criteria for the Diagnosis of Alzheimer’s Disease, a working group led by Bruno Dubois at Pierre and Marie Curie University, Paris, France, stepped into the void. The group comprises 24 clinician-researchers from five European countries, the U.S., Canada, and Japan, including—besides Pieter Jelle Visser and Stephen Salloway—Philip Sheltens of VU University Medical Center, Amsterdam; Howard Feldman of the University of British Columbia in Vancouver and Bristol Myers Squibb in Wallingford, Connecticut; and other leading AD clinicians in their respective countries. In essence, their criteria combine a deficit on an episodic memory test with an abnormal biomarker result as the twin requirements for a clinicobiological diagnosis of Alzheimer’s in the absence of dementia. In a subsequent October 2010 position paper in the Lancet Neurology, the group refined the terms they recommend the field use to diagnose the phases of AD.
Their lexicon defines two terms for the preclinical states of AD. People who are cognitively normal but have a positive brain amyloid PET scan or an AD-like signature in their cerebrospinal fluid (CSF) would be viewed as being “asymptomatic at risk for AD.” Asymptomatic people who are known to get AD in the future because they carry a rare autosomal-dominant AD mutation are labeled as having “presymptomatic AD.”
A little further down the road toward full-blown disease, people who are measurably impaired on selected memory tests, plus have an abnormal biomarker reading but no dementia, would get a diagnosis of “prodromal AD.” “Prodromal,” then, is already a symptomatic phase; “preclinical,” “asymptomatic at risk,” and “presymptomatic” are not. This is different from mild cognitive impairment (MCI), a term widely used in the U.S., in that a person diagnosed at the MCI stage would by Dubois criteria be recognized as having AD without yet having reached the disease’s dementia stage. Prodromal AD would not be predictive of AD, as is MCI, but would diagnose AD as the disease responsible for the episodic memory symptoms and the biomarker change. When a person subsequently develops dementia, this would not trigger the AD diagnosis but serve merely as a severity milestone separating the prodromal from the dementia phase. The MCI category, according to Dubois et al., would exclude people on this AD trajectory and be reserved as a clinical syndrome for the remaining, non-AD forms of cognitive impairment (Dubois et al., 2010). The lexicon paper also defines “Alzheimer’s disease dementia,” as well as atypical and mixed forms of AD.
In 2009, the NIA and the AA convened advisory meetings and established three working groups for the revision of the 1984 criteria, one each for AD, for its symptomatic pre-dementia phase, and its asymptomatic preclinical phase. The group leaders are, respectively, Guy McKhann, Johns Hopkins University, Baltimore, Maryland, Marilyn Albert, also there, and Reisa Sperling, Brigham and Women’s Hospital, Boston, Massachusetts. They released drafts of the new NIA/AA guidelines to the field and the press at the International Congress on Alzheimer’s Disease last July (see ARF related news story); a subsequent period of public commentary and refinement led to formal publication last April in three back-to-back papers and an introductory article coauthored by Clifford Jack of the Mayo Clinic in Rochester, Minnesota.
Beginning with the earliest phase of the disease, the working group led by Sperling gives biomarkers priority in identifying what the group calls “preclinical AD.” People in this situation would be cognitively normal or have very subtle cognitive decline that falls short of meeting criteria for MCI. They would also have a marker of “AD-pathophysiology (AD-P),” which the group distinguishes from “AD-clinical (AD-C).” The group states that AD-P can precede AD-C by many years, and that biomarker evidence of pathology greatly increases the risk of progressing (Sperling et al., 2011). As the disease progresses, different biomarkers become expressed, and the group exploits this timing to define three different preclinical stages. Stage 1 requires only a positive brain Aβ marker either by CSF or PET. Stage 2 requires stage 1 evidence plus signs of neurodegeneration, such as elevated CSF tau or functional or structural imaging. Stage 3 requires stage 1 and 2 evidence plus some sign of subjective cognitive deficit that is insufficient to meet criteria for MCI due to AD. In other words, the most advanced end of “preclinical AD” begins to express subtle clinical signs.
Sperling and colleagues note that it is possible that people with stage 1 or 2 preclinical AD may never progress further, but that those with stage 3 preclinical AD are likely to decline to MCI within several years. For trials, amyloid-modifying therapies might be less successful at stages 2 and 3, since these patients by definition have ongoing neurodegeneration, note the authors. On the other hand, they suggest that studies relying on cognitive outcomes, as do most pharmaceutical trials, might want to enrich for people with stage 3 preclinical AD, as these are expected to decline more quickly than those at stage 1 or 2.
Past the most advanced preclinical AD, people may become symptomatic, though not necessarily get dementia. The group led by Albert revised guidelines for the diagnosis of the symptomatic pre-dementia phase of AD (Albert et al., 2011). The group established the new entity “mild cognitive impairment due to AD” to distinguish it from MCI that might be due to other conditions. Albert and colleagues outline core clinical criteria for MCI due to AD, plus research criteria that incorporate biomarkers. The former stipulate that there should be concern over a decline in cognition and some evidence of impairment in one or more cognitive domains, typically including memory. For this diagnostic category to apply, both the concern and the evidence must occur absent dementia and with the person still able to stay functionally independent.
To this, the research criteria add biomarkers. The working group divided them into markers of Aβ and markers of neuronal injury. The former include CSF Aβ42 and PET brain amyloid; the latter include CSF tau and phosphorylated tau, structural and functional imaging such as brain atrophy and perfusion measures and FDG-PET. The criteria rank these markers hierarchically, whereby MCI due to AD is deemed “high likelihood” when both Aβ and neuronal injury markers test positive. A positive Aβ test in the absence of neuronal injury yields “intermediate likelihood,” as does a positive neuronal injury test in the absence of an Aβ test. If both tests are negative, then MCI is deemed “unlikely due to AD.” If biomarker data are conflicting—i.e., positive on Aβ, negative on neuronal injury—then biomarkers are deemed uninformative and diagnosis falls back to the clinical criteria.
Beyond MCI due to AD lies full-blown AD itself. The new NIA/AA guidelines for the diagnosis of AD (see McKhann et al., 2011) do not represent a dramatic change from what is currently in use in the clinic. The emphasis remains on a clinical diagnosis, but accounts for other forms of dementia that were not on the clinical radar in 1984, such as primary progressive aphasia (PPA) and frontotemporal dementia (FTD), and also for advances in genetics and biomarkers of pathophysiology, though the latter are only used to increase the certainty of clinical diagnosis. The panel recommends criteria for “all-cause dementia” and for “AD dementia” or “dementia caused by AD,” which the authors define as being secondary to the pathophysiology of AD. Criteria for all-cause dementia require cognitive or behavioral symptoms that interfere with work or usual activities, demonstrate a decline, are not explained by other causes, and require a combination of reported history and objective cognitive or neuropsychological testing in at least two domains.
The group proposed three possible diagnoses: 1) “probable AD dementia”; 2) “possible AD dementia”; and 3) “probable or possible AD dementia with evidence of the AD pathophysiological process.” The first applies when a person meets criteria for all-cause dementia with gradual onset and clear worsening. The most prominent cognitive deficit must be either in memory, language, visuospatial, or executive. A causative AD mutation increases the level of certainty of the “probable AD dementia” diagnosis, though McKhann and colleagues did not view ApoE4 as doing so. Possible AD dementia would apply when a person has core clinical criteria but “atypical course,” i.e., sudden onset or gaps in the evidence. Comorbidities such as cerebrovascular disease and evidence of FTD, PPA, or DLB suggest a diagnosis of “possible AD dementia” as well.
One important question on the minds of clinician-researchers, as well as other care providers, is whether the new criteria are meant for research purposes or for clinical use. Research criteria serve special uses such as drug testing, and their validation tends to pave the way for subsequent clinical criteria. Clinical criteria have to pass a higher bar in terms of robustness and accuracy, as they are applied in diverse community settings by a range of medical staff with varying expertise in AD. The Food and Drug Administration has formally qualified no AD biomarkers as yet. At this point, the Dubois criteria are generally said to be research criteria, and they are being used in some drug trials and in academic medical settings that conduct clinical research. The NIA/AA criteria are published as a mixture of clinical- and research-grade. That is, the clinical portions of the AD and MCI criteria are meant for use in clinical settings, but anything involving biomarkers and all of the preclinical AD criteria are for research use.
These fine distinctions notwithstanding, there is a lingering perception fueled in part by the media that patients across the country can expect their doctor to use these criteria before long (see, e.g., Cincinnati.com; Tulsa World). Other stories clearly state the research nature of the biomarker measures (e.g., Science Friday). Patients increasingly are asking their physicians for biomarker measurements they have heard about, and in truth, at least one of those measurements, structural MRI, is already being used in some non-research settings to support an AD diagnosis. So what exactly is an academic neurologist, a trialist, a private practice neurologist, a family medicine physician, to do?
In this Webinar, the panelists summarized both sets of new research guidelines to help you understand the differences as well as the significant overlap among them. Discussion explored whether the criteria will be adopted universally or differently in different countries.
University of Minnesota
After having listened to the Webinar and read the series of papers on NIA/AA and IWG diagnostic criteria, I find myself unclear on two points: 1) With respect to the NIA/AA criteria for MCI and AD, if biomarker testing is conducted, do the levels of certainty apply to the clinical diagnosis, or is the hierarchy of levels of certainty applicable only to research diagnoses?
In particular, if biomarkers from both domains are negative, does the clinical diagnosis change to "MCI (or dementia)—unlikely due to AD?" Using the NIA/AA criteria, should the lowest level of certainty of MCI due to AD be labeled "MCI-Core clinical criteria" or "MCI due to AD—Core clinical criteria?"
Guidelines Can Be Bad for Public Health
The diagnosis of mild cognitive impairment as a prodrome to Alzheimer's disease (MCI due to AD) will someday be an important milestone in the clinical application of neuroscience. Laboratories around the world are collaborating in determining the appropriate markers for prodromal AD, studying their individual and collective predictive power, and standardizing how the diagnostic tests should be conducted. But caution is in order. Testing procedures that work very well in highly specialized research settings may create unexpected problems when attempting the difficult leap to widespread clinical practice.
Great excitement, but also great confusion, surrounded last year's presentation of the NIA-AA clinical and research guidelines for MCI due to AD. I believe these guidelines in their early drafts were off the mark, and are overly optimistic even in their final version. Guidelines proposed by highly specialized researchers can be blind to broader clinical and public health risks they may pose. Researchers are understandably enthusiastic about their findings and eager to push the research and clinical envelope. This enthusiasm may make them relatively indifferent to how the guidelines can be misused in less skilled, less careful, and less scrupulous hands. In my experience, experts on any given disorder often make suggestions that are perfectly appropriate in their own research work, but are likely to lead to unintended consequences when translated to the wider clinical and public health world.
So here I offer a list of cautions that I believe deserve consideration as MCI due to AD matures into an official and widely used (and hopefully useful) clinical diagnosis.
1. It will probably take at least three to five years to standardize the emerging PET, hippocampal, and cerebrospinal measures to the point where they can leave the research lab and be used clinically to diagnose MCI. Wide-eyed media reports suggesting that clinical testing and effective treatment are imminent have been misleading. It is always better to under-promise and over-deliver than to over-promise and under-deliver. Press releases that accompany guideline revisions should emphasize limitations and cautions.
2. It is more likely that a pattern of test results (rather than any one single test) will be needed to accurately make the diagnosis.
3. This means that testing will be expensive. Funds allocated for clinical MCI testing will drain resources away from other clinical areas.
4. This is particularly true for widespread clinical testing (not research testing) now since there is currently no treatment for MCI. Although there is a pipeline of possible contenders, an effective treatment is unlikely to emerge for at least some years, perhaps more than a decade.
5. A rush to clinical testing at this stage would likely result in non-standardized, unreliable, and excessive testing with poor predictive power. Clinical patients drawn from the general population will be more difficult to diagnose accurately than highly selected research patients. Commercial labs guided by the profit motive will be less scrupulous in screening and testing procedures, in the frequency with which tests are repeated, and in their reporting of results. A battery of tests designed as a one-shot predictor may be repackaged (to increase volume and profits) as a useful means of following the progression of disease. If something can possibly be misused for profit by the commercial sector, guideline developers may as well assume that it will be.
This raises several important questions. Should guideline developers be considering the risks and costs of each guideline suggestion, not just its possible benefits? Should they feel concerned about how suggestions can be misused once they become part of routine clinical care? Should they worry that premature clinical use may result in misallocation of scarce health resources?
Personally, I believe guideline development needs to draw not just on the expertise of disease specialists, but also include epidemiologists, health economists, and public policy experts. The great difficulty of translating research results into clinical practice and health policy should be made explicit and transparent. Perhaps guidelines should come with a black-box warning of risks and limitations, of how suggestions can be misused in clinical practice to the detriment of the individual patient, and of how they may result in health resource misallocation.