Historical development

The concept of pre-dementia causing subsyndromal symptoms was described as early as 1962, when Kral^{Reference Kral11} described ‘benign senescent forgetfulness’. This encompassed mild fluctuating retrieval-based memory impairment, which he speculated could be a mild early form of senile atrophy that spared the Papez^{Reference Papez22} circuit. A shortcoming of Kral's description was that it lacked operational criteria, which can impede diagnostic reliability.^{Reference Spitzer, Endicott and Robins23}

Over 20 years later, Kral's concept was extended and operationalised by Crook and colleagues; they labelled their concept age-associated memory impairment (AAMI), which they defined as subjective memory complaint and objective memory impairment on a memory test at least one standard deviation below the mean for young adults.^{Reference Crook, Bartus, Ferris, Whitehouse, Cohen and Gershon24} By using healthy young adults as a reference sample, this definition lacked specificity, given that performance on psychometric tasks of memory declines with healthy ageing^{Reference Wechsler25–Reference Mitrushina, Boone, Razani and D'Elia27} and up to 90% of elderly individuals would fulfil this criterion.^{Reference Smith, Ivnik, Petersen, Malec, Kokmen and Tangalos28} A further criticism was that AAMI exclusively focused on memory, although other cognitive domains, such as visuospatial abilities, language or executive functions, may be affected principally early in Alzheimer's disease.^{Reference Galton, Patterson, Xuereb and Hodges29,Reference Johnson, Head, Kim, Starr and Cotman30}

Addressing both of these criticisms, the International Psychogeriatric Association broadened the concept to include other cognitive domains and also defined objective impairment with reference to an age-matched sample.^{Reference Levy31} They labelled this age-associated cognitive decline (AACD), defined as subjective cognitive decline as observed by the individual or an informant; gradual decline over at least 6 months; and impairment in a cognitive domain with performance one standard deviation below the mean of an age- and education-matched normative sample.^{Reference Levy31} AAMI and AACD appear to be distinct clinical entities with only approximately 50% overlap in concordant diagnosis and AACD participants showing more extensive cognitive impairment.^{Reference Richards, Touchon, Ledesert and Richie32}

The term ‘MCI’ was first described by Reisberg and colleagues with the development of the Global Deterioration Scale.^{Reference Reisberg, Ferris, de Leon and Crook33} This was a seven-point ordinal scale from ‘no cognitive decline’ to ‘severe dementia’ that defined MCI as one or more of several examples of cognitive lapse such as becoming lost in an unfamiliar location, word-finding difficulty, forgetting names or misplacing objects, or as concentration deficit with clinical testing.^{Reference Reisberg, Ferris, de Leon and Crook33}

Concurrently, the clinical dementia rating (CDR) scale was developed,^{Reference Hughes, Berg, Danziger, Coben and Martin34} which was also an ordinal scale ranging from ‘no impairment’ to ‘severe dementia’. Although not directly referring to MCI, the CDR introduced the importance of daily functioning into the concept. A person scoring 0.5 or ‘questionable impairment’ on the CDR may have slight impairment of community affairs or home life but would be fully independent with self-care.^{Reference Hughes, Berg, Danziger, Coben and Martin34} Flicker and colleagues used the term ‘MCI’ when they showed that psychometric impairment at baseline could predict subsequent decline in elderly patients after 2 years.^{Reference Flicker, Ferris and Reisberg35}

Ronald Petersen, a major developer of the concept of MCI through the Mayo clinic, developed his original definition of MCI based on patients recruited from a community-based medical clinic.^{Reference Petersen, Smith, Ivnik, Tangalos, Schaid and Thibodeau36} They identified people who were themselves concerned about their cognition, or whose carers or physicians were concerned. These patients then had an extensive battery of physical examination, cognitive assessment, investigations and neuroimaging to rule out dementia as determined by expert panel consensus. These patients by definition had ‘normal’ scores on the Mini-Mental State Examination^{Reference Folstein, Folstein and McHugh37} and Short Test of Mental Status.^{Reference Kokmen, Naessens and Offord38} Petersen and colleagues (1995) observed that this cohort tended to perform 1.5 standard deviations below the age-matched mean performance on memory tasks such as auditory verbal learning tests, and activities of daily living (ADL) were generally preserved, corresponding to a CDR rating of 0.5. By employing age-corrected, but not education-corrected, normative data, it introduced confounding difficulties with patients with low education or low IQ.

These criteria were more formally proposed and became known as the Mayo Clinic core criteria or the Petersen criteria.^{Reference Petersen, Smith, Waring, Ivnik, Tangalos and Kokmen39} The criteria were restricted to memory impairment rather than impairment of other cognitive domains, and thus were subject to similar criticism to that of AAMI; that Alzheimer's disease may principally affect other cognitive domains.^{Reference Galton, Patterson, Xuereb and Hodges29} In 2003, a key symposium of experts revised the Mayo Clinic criteria to include domains other than memory.^{Reference Winblad, Palmer, Kivipelto, Jelic, Fratiglioni and Wahlund12} Referred to as the Winblad criteria, these defined MCI as: (a) the person is neither normal nor demented; (b) there is evidence of cognitive deterioration shown by either objectively measured decline over time and/or subjective report of decline by self and/or informant in conjunction with objective cognitive deficits; and (3) ADL are preserved and complex instrumental functions are either intact or minimally impaired.^{Reference Winblad, Palmer, Kivipelto, Jelic, Fratiglioni and Wahlund12} Subcategories of MCI were established based on the pattern of cognitive domains affected: amnestic single-domain, amnestic multiple-domain, non-amnestic single-domain and non-amnestic multiple-domain.^{Reference Petersen40}

The concept of ‘cognitive impairment, no dementia’ (CIND) was introduced in the context of the need for early recognition of dementia.^{Reference Graham, Rockwood, Beattie, Eastwood, Gauthier and Tuokko41} CIND was identified on the basis of a consensus conference of physician, nurse and neuropsychologist, integrating all available information from clinical and psychometric assessment.^{Reference Chertkow, Nasreddine, Joanette, Drolet, Kirk and Massoud42} It includes individuals with non-dementia-related aetiologies such as delirium, chronic alcohol and drug use, depression, psychiatric illness, intellectual disability and circumscribed memory impairment; this results in high prevalence estimates^{Reference Graham, Rockwood, Beattie, Eastwood, Gauthier and Tuokko41} and many CIND individuals will not develop dementia. A criticism of CIND is that it does not provide operational criteria, which may jeopardise its reliability.

Recent definitions and developments

In the context of emerging biomarkers, the National Institute on Aging and the Alzheimer's Association (NIA-AA) convened a workgroup to revise the diagnostic criteria for pre-dementia Alzheimer's disease.^{Reference Albert, DeKosky, Dickson, Dubois, Feldman and Fox43} Not long thereafter, the DSM-5¹ abandoned the term ‘dementia’ and replaced it with ‘major neurocognitive disorder’, while adding the term ‘mild neurocognitive disorder’ (mNCD), which has similarities to MCI including cognitive complaint, psychometric impairment and relative preservation of ADL.

The NIA-AA and DSM-5 mNCD both refrained from offering a strict cut-off score for psychometric impairment, instead suggesting that typical levels of impairment would be 1–2 or 1–1.5 standard deviations below the mean, respectively, for age- and education-matched normative data. Instead of arbitrary cut-offs, these criteria advocated for an individualised assessment that incorporated all available evidence.

The NIA-AA criteria^{Reference Albert, DeKosky, Dickson, Dubois, Feldman and Fox43} combined core clinical criteria with clinical research criteria, which incorporated biomarker evidence of disease. In doing so, these criteria moved beyond MCI as a pre-clinical definition incorporating history and examination findings to a prodromal state with biological evidence of incipient disease. The NIA-AA workgroup explicitly focused on MCI due to Alzheimer's disease and used biomarkers to stratify the likelihood that the cognitive change is due to Alzheimer's disease. Biomarkers indicating a high likelihood that MCI is due to Alzheimer's disease are an abnormal Aβ marker (e.g. positive PiB (Pittsburgh compound B) scan or cerebrospinal fluid (CSF) Aβ₄₂) and a positive biomarker of neuronal injury (e.g. CSF tau, FDG-PET (Fluorodeoxyglucose Positron Emission Tomography) or structural magnetic resonance imaging). MCI unlikely to be due to Alzheimer's disease is determined when Aβ markers and markers of neuronal injury are both negative. MCI due to Alzheimer's disease with intermediate likelihood has either Aβ markers or neuronal injury markers as abnormal, while the other is untested. Recent studies suggest this taxonomy is useful in predicting Alzheimer's disease.^{Reference Vos, Verhey, Frolich, Kornhuber, Wiltfang and Maier44,Reference Guo, Alexopoulos, Eisele, Wagenpfeil, Kurz and Perneczky45} However, the invasiveness, cost and availability of these biomarkers may limit their widespread implementation in clinical settings.

Subjective cognitive decline is incorporated into modern definitions of MCI.^{1,Reference Winblad, Palmer, Kivipelto, Jelic, Fratiglioni and Wahlund12,Reference Crook, Bartus, Ferris, Whitehouse, Cohen and Gershon24,Reference Levy31,Reference Petersen, Smith, Ivnik, Tangalos, Schaid and Thibodeau36,Reference Albert, DeKosky, Dickson, Dubois, Feldman and Fox43} MCI may be preceded by a state in which the individual experiences subjective cognitive decline that is too subtle to be detected on psychometric testing.^{Reference Jessen, Amariglio, Van Boxtel, Breteler, Ceccaldi and Chételat46} The Subjective Cognitive Decline Initiative working party have conceptualised this as a pre-MCI state on the same spectrum towards Alzheimer's disease dementia and provided definitions^{Reference Jessen, Amariglio, Van Boxtel, Breteler, Ceccaldi and Chételat46} that have been operationalised for research purposes.^{Reference Molinuevo, Rabin, Amariglio, Buckley, Dubois and Ellis47}

Prevalence of MCI

Since their publication, the revised Mayo clinic criteria^{Reference Winblad, Palmer, Kivipelto, Jelic, Fratiglioni and Wahlund12} have been commonly adopted in the literature, and studies reported in this section used these criteria unless otherwise stated.

Prospective population-based studies show that the prevalence of MCI ranges from 15 to 22% in elderly individuals.^{Reference Lopez, Jagust, DeKosky, Becker, Fitzpatrick and Dulberg48,Reference Petersen49} Prevalence increases with age, decreases with education, and is more common in males, unmarried people and carriers of the APOE-ε4 allele.^{Reference Petersen49,Reference Petersen, Roberts, Knopman, Geda, Cha and Pankratz50} Prospective population-based studies have estimated incidence rates of around 6% per year, although the rate in men (over 7%) was slightly higher than that in women (under 6%).^{Reference Roberts, Geda, Knopman, Cha, Pankratz and Boeve51}

Progression of MCI to dementia

Estimates of progression rates to dementia or Alzheimer's disease dementia are important for advising patients about prognosis and have implications for conducting research in this population. Individuals with MCI have a higher risk of developing dementia compared with the general older population incidence of 1–2% per year,^{Reference Petersen, Stevens, Ganguli, Tangalos, Cummings and DeKosky52} although estimates vary depending on the definition or subtype of MCI, study design and follow-up period.^{Reference Petersen, Stevens, Ganguli, Tangalos, Cummings and DeKosky52–Reference Petersen, Thomas, Grundman, Bennett, Doody and Ferris55} Earlier definitions using the Petersen amnestic-only MCI criteria estimated rates of progression to Alzheimer's disease dementia to be 10–15% per year.^{Reference Petersen, Stevens, Ganguli, Tangalos, Cummings and DeKosky52} A randomised controlled trial reported a progression rate of 16% per year.^{Reference Petersen, Thomas, Grundman, Bennett, Doody and Ferris55} A meta-analysis of studies using Mayo clinic criteria for MCI suggested that over 10 years, 33.6% will cumulatively progress to Alzheimer's disease dementia in specialist settings versus 28.9% in population settings, which translated to an annual progression rate of 8.1% in specialist settings and 6.8% in community studies.^{Reference Mitchell and Shiri-Feshki56}

There is some criticism of the utility of MCI as a diagnosis given its heterogenous nosology,^{Reference Werner and Korczyn57} variable prognostic significance^{Reference Bowen, Teri, Kukull, McCormick, McCurry and Larson58–Reference Ganguli, Dodge, Shen and DeKosky60} and the various ethical issues it raises.^{Reference Werner and Korczyn57} We would counterargue that these issue provide impetus to refine the definition of MCI, as doing so will allow identification of a group that could be identified for treatment of modifiable risk factors that may decrease the risk of developing dementia, such as diet, diabetes mellitus, hypertension and hypercholesterolemia.^{Reference Cooper, Sommerlad, Lyketsos and Livingston61,Reference Li, Wang, Zhang, Xu, Gao and Fang62}

The present study aimed to review the evidence with regards to which taxonomy of MCI was more useful in predicting incident Alzheimer's disease dementia. We hypothesised that amnestic MCI (aMCI) and multiple-domain MCI would be more likely than non-MCI controls to progress to Alzheimer's disease dementia.