Article Text

Download PDFPDF

Temporal trends in the long term risk of progression of mild cognitive impairment: a pooled analysis
  1. A J Mitchell1,
  2. M Shiri-Feshki2
  1. 1
    University Hospitals Leicester and University of Leicester, Leicester, UK
  2. 2
    Nottinghamshire Healthcare NHS Trust, Nottingham, UK
  1. Dr A J Mitchell, University of Leicester, Brandon Unit Leicester General Hospital, Leicester LE5 4PW, UK; alex.mitchell{at}


Background: Mild cognitive impairment (MCI) is a condition that carries a substantial risk of dementia. The exact magnitude of that risk is uncertain because of the variations in the definition of MCI, the setting (such as memory clinic versus community) and, equally importantly, the duration of follow-up. Recently, a number of long term studies have been published with observation periods of 5 years of longer.

Methods: In this quantitative review, 15 long term studies were examined and compared with the results from studies using shorter periods of observation, focusing on the annual conversion rate (ACR) of MCI to dementia.

Results: The report identified six long term clinical studies conducted in specialist settings and nine long term population studies conducted in the community. Across all cohort studies with completed follow-up, the mean ACR to dementia was 4.2% (95% CI 3.9% to 4.6%). This was lower than the rate reported in studies of less than 5 years’ duration. The cumulative conversion rate averaged 31.4% in this sample. The proportion converting to dementia (and Alzheimer’s disease) declined with longer observation periods, suggesting that risk of progression diminishes with time.

Conclusion: A 10–15% ACR only appears to hold true in clinical samples monitored over a short observation period. Recruitment of older individuals from specialist centres, particularly involving those who complain of cognitive difficulties (subjective memory complaints) will tend to favour high conversion rates. In the first few years of follow-up, many of those with the most adverse risk profile will tend to progress, dropout or die, leaving a cohort of less vulnerable sufferers. However, an inverse temporal relationship was also manifest in those who completed long term follow-up, suggesting other factors are involved such as sampling issues or heterogeneity in MCI itself.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

The first operational definition of mild cognitive impairment (MCI) was suggested in 1991 by Flicker et al at New York University.1 It was redefined shortly thereafter by Zaudig (1992) at the Max Planck Institute2 and again by Petersen et al (1997 and 1999) at the Mayo Clinic.3 4 None of these definitions has been universally accepted, leading to controversy about optimal criteria.5 6 An attempt to reach consensus has been proposed recently by the MCI Working Group of the European Consortium on Alzheimer’s disease.7 The core concept is of verified cognitive impairment that is of insufficient severity to warrant a diagnosis of early dementia or to cause substantial function impairment but usually noted by an individual or relative.7 8 Clinicians may apply informal addition exclusions, for example that other explanations for cognitive impairment such as depression are ruled out. Regardless of exact definition, the clinical utility of MCI is fundamentally its ability to predict later dementia.9 Numerous short term studies have generated a view that the annual conversion rate (ACR) is 10–15%.10 If this annual conversion rate held true, then within 10 years of diagnosis almost all MCI suffers would have developed dementia. Indeed, Peterson commented that after 6 years, 80% of the Mayo Clinic MCI cohort had progressed to dementia and he further estimated that the final proportion might extend to 90%.11 12 He was not alone in this assertion as other groups documented a 44–64% conversion rate within 2 years of initial observation.1 1315 This has led many to suggest that MCI is an inescapable intermediate stage between normal aging and dementia.16 Yet it is well known that many patients with MCI do not deteriorate and some patients improve over time.1719 For example, Wolf et al demonstrated that over almost 3 years, 20% of MCI sufferers had recovered and an additional 60% neither improved nor deteriorated.17 An alternate possibility therefore is that MCI may not be a homogenous condition but may comprise several disease groups united by propensity to cause modest cognitive impairment. If this hypothesis were true, one might expect considerable variation in the long term natural history and response to new treatments in some but not all sufferers.

Given that MCI as a defined concept is only 15 years old, it has taken until now to accrue sufficient prospective data to allow statements of prognosis to be stated with any kind of confidence. In the past few years a large number of cohort studies with follow-up durations that exceed 2 years have attempted to answer the question, what is the long term rate of conversion from MCI to dementia? While the maximum follow-up for some individuals defined retrospectively can extend to 20 years, it is the mean follow-up period in larger studies that is the most informative.20 The two longest studies published to date both had mean observation periods of 10 years. Visser et al reported the long term risk of dementia (and Alzheimer’s disease) in subjects with MCI aged 40–85 years from a memory clinic sample.28 The ACR to dementia in this study was 4.8%. Ganguli et al also conducted a 10 year study but based on a community sample.34 Of those retrospectively diagnosed with MCI, 27% developed dementia over the next 10 years, an ACR of 2.75% but 55.0% of those with MCI no longer met criteria for either MCI or dementia by the time of the third wave of follow-up (approximately 6 years). The largest study looked at nuns and brothers in the US, of whom 949 had evidence of MCI at baseline. Over an 8 year period, only 7.5% developed dementia and almost 15% recovered.38 This equates to a mean ACR of 0.94%.

Given the remarkably low ACR in these robust studies, we aimed to examine systematically all long term MCI studies, each with a minimum of 5.0 years of longitudinal observation, analysed as a proportion of those recruited at baseline (inception cohort method) and using all those present at follow-up (completer method).



A systematic search and pooled analysis was conducted for studies examining the progression of MCI for 5 years or longer. The following abstract databases were searched: Medline/Pubmed 1966–March 2008, PsycINFO 1887–March 2008 and Embase 1980–March 2008. The following search terms were used: “progress* or convert* or prognosis or deteriorate or develop or decline* or cohort or longitude* or prospective* [Abstract]” AND “dementia or Alzheimer* or mild cognitive [Abstract]”.

Inclusion criteria

Authors’ definitions of MCI were carefully recorded (eg, those that adhered to the original 1997 Petersen et al criteria for MCI3 or the revised Petersen et al criteria16). There were insufficient data to examine amnestic compared with non-amnestic subtypes. We excluded studies that defined cognitive impairment solely on the basis of age associated memory impairment,30 cognitively impaired not demented21 22 or Clinical Dementia Rating severity (Clinical Dementia Rating is 0.5 or questionable dementia).2326 Regarding outcome measures, we allowed any operational (such as International Classification of Diseases (ICD)-10, Diagnostic and Statistical Manual of Mental Disorders version III revised (DSM-IIIR) or DSM-IV) or neuropsychologically based definition of dementia or probable Alzheimer’s disease.

Statistical analysis

Unweighted (raw) rates of progression were calculated as a proportion of those recruited at baseline (inception cohort method) and also using all those present at follow-up (completer method). The completer method excludes cases who were lost to follow-up and who died before study completion. As we did not have access to the raw data of each case, a Kaplan–Meyer survival analysis was not possible. Regression was performed with StatsDirect using linear regression analysis by using the “least squares” method. Relationship between one or more study level characteristics can be compared using meta-regression.

The inception annual conversion rate  = 

(number of cases of dementia that developed during observation/number of MCI patients at recruitment)/mean years of observation.

The completer annual conversion rate  = 

(number of cases of dementia that developed during observation/number of cases present at end of observation)/mean years of observation.


We identified 15 long term cohort studies that counted the number of subjects recruited at baseline. These had a mean observation period of 6.02 years (range 5.0–10.0 years) (table 1). One study was excluded because of likely duplication of data.27 Six took place in specialist clinical settings, such as memory clinics,2833 and nine were population based (in the community).3442 Of 15 studies, 10 reported generic dementia as an outcome and 11 reported Alzheimer’s disease.

Table 1 Summary of cohort studies examining the long term progression of mild cognitive impairment to dementia

Cumulative progression

The cumulative conversion rate to dementia averaged 31.4% in these long term studies. The cumulative conversion to Alzheimer’s disease was 32.8%. Only one study demonstrated a progression rate above 50% to dementia33 and no study showed a greater than 50% conversion to Alzheimer’s disease. Compared with studies of less than 5 years duration where the conversion rate to dementia and Alzheimer’s disease was 27.4% and 30.5% (data available on request), there was no significant difference in cumulative conversion in long compared with short term studies.

Annual rate of progression

Across all 15 long term studies, the mean ACR to dementia was 3.3% (95% confidence interval (CI) 3.0% to 3.6%) (table 2). In long term studies from specialist settings, the ACR to dementia was 6.7% (95% CI 5.9% to 7.5%) compared with 2.6% (95% CI 2.3% to 2.9%) for those conducted in the community, a highly significant difference (p<0.01). If analysis was limited to studies that defined Alzheimer’s disease as the main outcome, then the ACR was 5.1% (95% CI 4.7% to 5.6%, n = 11). Again, there was a significantly higher conversion rate to Alzheimer’s disease in studies from specialist rather than community settings (p<0.01).

Table 2 Summary of annual conversion rates from mild cognitive impairment to dementia

Analysing the data using the completer cohort method, the mean ACR to dementia was 4.2% (95% CI 3.9% to 4.6%) and the ACR to Alzheimer’s disease was 5.8% (95% CI 5.5% to 6.5%). The completer conversion to dementia was significantly higher than the inception cohort conversion rate (p<0.05). There was no significant association with age and the ACR to dementia by inception cohort or completer cohort methods on meta-regression.

Temporal trends

We analysed the association of ACR with duration of observation of study. There was a significant inverse relationship in both inception studies (r = 0.59, p = 0.02) and completer studies (r = 0.53, p = 0.04). The relationship was also significant where the outcome was Alzheimer’s disease alone (p<0.05). The best fit was exponential with a lower ACR the longer the period of follow-up (fig 1). Compared with the ACR of 7.8% in inception studies of 3.0–4.9 years, the relative risk of annual progression was 0.42 in long term studies.

Figure 1 Annual conversion rate to dementia (generic or Alzheimer’s disease) in 15 long term studies, according to the length of observation.


The widely cited 10–15% conversion rate cannot be extrapolated indefinitely from MCI studies with short observation periods. Indeed, only one long term study showed an ACR above 10%.33 The mean annual conversion rate from 15 studies that exceeded 5 years of follow-up was only 3.3% (5.1% to Alzheimer’s disease), much lower than that previously suggested, and subject to an inverse temporal relationship. We are not the first to suggest that the conversion rate during the first year of observation does not hold linearly thereafter, even within the same cohort.4345

The highest rates of conversion were often seen in clinical samples recruited from specialist centres, such as memory clinic samples. These results are consistent with previously findings. For example, Larrieu et al for The Personnes Agees QUID study (PAQUID) followed a community based cohort of healthy elderly subjects for 5 years.36 The ACR rate was 9.0% among those with MCI but with considerable instability. Within 3 years, only 6% of the subjects continued to have MCI and about 40% reverted to normal. A similar finding was reported from the Kungsholmen Project based on the Swedish population aged 75 years and above.46 47 After 3 years, about one-third of persons with baseline cognitive impairment progressed to dementia but 25% of subjects improved and generally remained well. Of note, the relative risk of progressing to dementia was higher over the first 3 years than over the subsequent 3 years. This was replicated by Busse et al who studied a community sample of 1045 dementia-free individuals aged 75 years in the three wave longitudinal Leipzig Longitudinal Study of the Aged.48 After 6 years of observation, 20% of participants with MCI at baseline were still alive without a diagnosis of dementia and about 50% had developed dementia.35 Progression from MCI to dementia during the first 18 months of observation was about 20%, reducing to 10% after further follow-up.49 Visser et al reported the 10 year risk of dementia in subjects with MCI aged 40–85 years from a memory clinic sample.28 The majority of subjects with MCI did not progress to dementia, even after 10 years, and this effect was strongly moderated by age. The annual conversion rate was highest during the first year of follow-up and decreased with time. It was 10.8% during the first 2 years of follow-up, 4.5% during the next 3 years and 2.5% during the last 5 years. The largest sample to date is from the Three City community study involving 2882 subjects with MCI followed for 4 years.50 Only 6.6% progressed to dementia, 56.5% remained stable and 37% improved.

An important question then is why might the annual risk of dementia diminish with time? Numerous risk factors for conversion have been discovered, including atrophy of medial temporal lobe, CSF biomarkers, Apo E4 and neuropsychological markers.33 51 52 However, there is nothing to suggest these risk factors diminish substantially with time, indeed the effect of age obviously increases, making this effect more remarkable. A more plausible hypothesis is that any selection of MCI sufferers will include a heterogeneous group with varying neuropathologies or indeed no neuropathology at all. It follows that those with aggressive conditions will convert early, leaving a surviving cohort with relatively high resilience. It is also likely that many surviving cases lost to follow-up would have developed dementia with time. Thus inception studies may be subject to an attrition effect whereby lost cases influence temporal trends. However, our data show a reduction in annual progression risk even if non-completers are removed. An additional factor might be diagnostic error. If some cases of early dementia are mislabelled as MCI (an area of considerable diagnostic difficulty) and such individuals will deteriorate relatively quickly. This may occur in clinical and population studies. Memory clinics are high prevalence settings in which attendees have noted subjective cognitive complaints and other risk factors for progression. This helps explain why conversion rates for memory clinic samples were significantly higher than community samples.

These data on the longitudinal course of MCI suggest that MCI is not a unified disease but a heterogeneous disorder which may be stable or transitional. The cumulative conversion rate from the longest five studies was approximately 20–30% which is certainly a cause for concern but not as high as originally suggested. This is an important message for newly diagnosed individuals and families who may struggle to cope with an imprecise diagnosis.53 Trials designed to modify the course of MCI may need to be re-thought in light of these findings.54 Future work should aim to further subtype MCI not simply into fast and slow decliners but into those destined to remain stable, decline or even improve.55 56


Thanks to the staff of the postgraduate library, Leicester General Hospital. Authors’ contributions: MSF (literature search, critical appraisal, writing) AJM (all tasks).


View Abstract


  • Competing interests: None.