Article Text

Download PDFPDF
Early intervention for mild cognitive impairment: a randomised controlled trial
  1. G J Kinsella1,2,
  2. E Mullaly2,3,
  3. E Rand3,
  4. B Ong1,
  5. C Burton1,
  6. S Price1,
  7. M Phillips4,
  8. E Storey3,5
  1. 1
    Psychological Science, La Trobe University, Melbourne, Australia
  2. 2
    Psychology Service, Caulfield Hospital, Melbourne, Australia
  3. 3
    Cognitive Dementia and Memory Service, Caulfield Hospital, Melbourne, Australia
  4. 4
    Cognitive Dementia and Memory Service, Bundoora Extended Care Centre, Melbourne, Australia
  5. 5
    Department of Neuroscience (Medicine), Monash University (Alfred Hospital Campus), Melbourne, Australia
  1. Professor G Kinsella, Psychological Science, La Trobe University, Melbourne, Victoria 3086, Australia; g.kinsella{at}


Background: Positive effects are reported for memory training for healthy older adults, and yet there is limited information about the benefit of cognitive intervention for older adults with increasing memory difficulties—mild cognitive impairment.

Objective: To investigate the usefulness of an early cognitive intervention for the memory difficulties experienced by people with amnestic mild cognitive impairment.

Methods: Using a randomised control design, 52 participants with amnestic mild cognitive impairment and their family partners were randomly assigned to a cognitive intervention (memory rehabilitation group) or waitlist (control group). Participants were assessed on primary measures of everyday memory (prospective memory) and memory strategies at 2 weeks’ and 4 months’ follow-up; secondary measures of contentment with memory and the family participants’ knowledge of memory strategies were also assessed.

Results: Everyday memory, measured by performance on prospective memory tasks, significantly improved following intervention, although self-appraisal of everyday memory did not demonstrate a similar intervention effect. Knowledge and use of memory strategies also significantly increased following intervention. Furthermore, family knowledge of memory strategies increased following intervention. There was a strong trend towards improvement in contentment with memory immediately following intervention, but this effect was not significant.

Conclusions: Early intervention for memory difficulties in amnestic mild cognitive impairment, using cognitive rehabilitation in compensatory strategies, can assist in minimising everyday memory failures as evaluated by performance on prospective memory tasks and knowledge of memory strategies.

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.

Many older adults reviewed at memory clinics present with a report of memory impairment that can be detailed on neuropsychological assessment but fails to reach accepted criteria for a clear diagnosis of Alzheimer disease (AD).15 The memory deficit occurs in the context of preserved general cognition and functional ability in basic everyday tasks, and is referred to as mild cognitive impairment (MCI) or, more specifically, amnestic MCI (aMCI—single or multiple domains).48 The importance of aMCI is that progression to AD, or at least to non-specific dementia,9 has been reported to occur at a rate of 12–15% per year as distinct from 1–2% in healthy adults.4 10 11

Given the increased risk of developing AD in this population, there is a strong argument for developing cognitive interventions to assist with memory problems. Pharmacological treatment of early dementia is currently under active development,12 and cognitive intervention, if found to be successful at this early stage, can provide added value to pharmacological management. Furthermore, the earlier that compensatory behavioural strategy can be introduced to persons struggling with declining memory, the more likely it will be that memory strategies will be used effectively in everyday life.13

The value of memory training for healthy older adults was addressed in an early meta-analysis14 that reported significant prepost gains by trained groups. More recently, several randomised trials have further investigated the benefits of memory training in older adults,15 16 and positive effects were reported even after a 5-year follow-up.17 However, there has been very limited study of the response of persons with aMCI to memory intervention, and there is no clear consensus about the extent to which participants with aMCI can benefit from cognitive training1822

These studies encourage further investigation,23 especially as the direct benefit on daily functioning has not been evaluated. This is important, as the memory deficit in aMCI is more problematic than for healthy older adults, and a critical issue is to improve the memory disability engendered in everyday activities rather than memory impairment per se as indexed by performance on memory tests. A common view in neurological rehabilitation is that change in the underlying impairment of function is difficult to achieve, considering limitations in neuronal plasticity; however, resultant disability, that is the impact of impairment in everyday function, can be moderated by training in compensatory strategies.24 Troyer et al22 have provided support for this approach by demonstrating that participants with aMCI can be taught knowledge and use of strategies but also argued that the impact of these observed changes needs to be demonstrated in everyday behaviour. Many everyday activities rely on prospective memory, or remembering to perform intended actions at the appropriate time, and complaints among older adults are most frequently about prospective memory errors (eg, forgetting to turn off a hose pipe).25 Consequently, prospective memory can provide a useful approach to measurement of everyday memory competence.

Finally, by integrating families into interventions designed for patients with aMCI, it is anticipated that this will provide foundations of support as families move to a greater caregiving role in response to the evolution of AD.26 Furthermore, early intervention to provide management skills for the family is also advocated, as it has been demonstrated that within an AD population, care-giver management strategies influence patient behaviour.27

Therefore, our aim was to evaluate an early cognitive intervention for older adults diagnosed as having aMCI. Rather than attempting to change the underlying memory impairment, our primary objective was to minimise memory disability (prospective memory errors) by increasing awareness of memory issues in everyday life and by using specific strategies to prevent memory failures. Importantly, the intervention involved the family, as it is recognised that families are often required to become care givers. We expected that following intervention participants with aMCI would demonstrate improved everyday memory performance (through better knowledge and use of memory strategies), especially in prospective memory (primary outcomes). We also expected that family participants would demonstrate improved knowledge of memory strategies, and participants with aMCI would report improved psychological well-being (secondary exploratory outcomes).



Ethics approval was obtained from La Trobe University and participating hospitals, and the aMCI sample was recruited from people presenting with a memory complaint at two memory clinics in Melbourne between October 2004 and January 2007. The diagnostic process of the memory clinics was based on a comprehensive assessment including interviews with patient and family, laboratory and radiological studies, and neurological, psychiatric and neuropsychological examination. Inclusion criteria for aMCI35 were: (i) subjective memory complaint by patient and/or an informant; (ii) objective memory impairment on neuropsychological tests of memory (ie, Hopkins Verbal Learning Test—Revised,28 Rey Auditory Verbal Learning Test,29 Wechsler30 Logical Prose Passages, Word List Learning or Verbal Paired Associates)—1.5 SD or more below age/education norms; (iii) relatively normal performance in general cognition based on structured interview with the patient and informant and the Mini-Mental State Examination (MMSE31); (iv) no impairment in personal activities of daily living (P-ADL) as determined by clinical interview with participant and family (instrumental activities of daily living, IADL, may be minimally impaired); and (v) failure to reach criterion for clinical dementia—National Institute of Neurological Communicative Disorders and Stroke and the Alzheimer disease and Related Disorders Association (NINCDS-ADRDA)32 guidelines for dementia. Exclusion criteria were as follows: (i) comorbid medical conditions associated with functional decline or cognition; (ii) major psychiatric disorder and/or behavioural problems; (iii) significant cerebro-vascular disease; (iv) significant impairment of vision, hearing or communication. Using these criteria, potential participants developed aMCI (single- or multiple-domain), and were further screened by the research team for predicted IQ (Wechsler Test of Adult Reading, WTAR33), general cognition (Mini-Mental State Examination, MMSE31), objective memory impairment (Hopkins Verbal learning Test—Revised, HVLT-R28 34), sufficient conversational English and available time for group participation.

Study design and procedure

A randomised controlled, single-blind trial evaluated the impact of a 5-week memory group intervention, using a two-group design—an intervention and a waitlist control group. Participants were recruited in cohorts from two memory clinics when there were sufficient referrals to form an intervention and control group (at least eight families). At recruitment, families were offered a training programme to learn about coping with memory difficulties. It was explained that there was a waiting list, and although there may be an opportunity to participate in a programme in several weeks, there may be a delay of several months. If willing to participate, informed consent procedures were undertaken, and the participant was randomly assigned to either group by a researcher not involved in the assessments or intervention, using a random table blocked into group sizes of eight. Families in the intervention group were invited to the next intervention scheduled to start within a couple of weeks, and families in the waitlist group were informed that there would be a delay until they could join the intervention programme. However, in order to initiate the programme, participants in the first cohort were all allocated to the intervention group.

Assessments were undertaken at baseline, 2 weeks after intervention and at 4-month follow-up by a research clinical neuropsychologist masked to treatment group.


Memory intervention was conducted by experienced health professionals (clinical neuropsychologists and an occupational therapist), and consisted of five weekly 1½ h sessions, located in hospital-based outpatient memory clinics. Sessions used a problem-solving approach to illustrate common everyday memory problems and practice in strategies to respond to these problems. Written session material was provided to each family, and assignments were undertaken to facilitate active application of the skills in everyday activities.

The methods selected for the intervention are reflective of current scientific research in memory processing, that is evidence-based.15 22 25 35 The first session provided a discussion of memory as a multifactorial construct and elaborated how everyday memory changes can be experienced as a result of health and lifestyle issues, ageing or neurological disorder. As the most frequently used approach to compensation of memory impairment is the use of an external aid,25 sufficient time was provided in the following sessions to explore the variety of external memory aids that can be considered for various tasks and situations. Strategies for improving organisational and attention skills in approaching learning and remembering were practised.25 Easily implemented memory strategies that can be used for specific everyday tasks (eg, face–name recall or remembering to do things) were also practised—verbal categorisation and elaboration,36 visual imagery,35 errorless learning35 and spaced retrieval.37 38 In addition, general coping strategies39 were discussed, including the value of physical and mental exercise.17 Throughout the programme, it was highlighted that successful strategies for everyday tasks frequently need to be combined, and different strategies will be more or less effective for different individuals.


Memory in everyday activities: performance and self-report

Prospective Memory Performance—(i) Reminding Task (modified from the Rivermead Behavioural Memory Test40)

At the start of testing, the details of the next appointment were written on an appointment card. The participant was requested to remind the assessor at the end of session to provide the appointment card. Scoring was follows: 2 points if request made within 15 s of testing completion, 1 point if request made late or required prompting and 0 point if participant could not recall instructions.

Prospective Memory Performance—(ii) Envelope Task41

Participants were instructed that later on they would be dictated a name and address to write on the front of an envelope. They were also to remember, without prompting, to write their name and initials on the back of the envelope and seal it. After a 10 min delay, the envelope was presented and the address dictated. Scoring was follows: 2 points if the name and initials were written within 5 s without prompting, 1 point if either the name or initials were written late or not performed and 0 point if no actions performed. These two tasks of prospective memory were combined to provide an index of memory ability/disability (possible range of scores 0–4; higher scores represent more memory ability).

Multifactorial Metamemory Questionnaire—Ability subscale (MMQ-Ability42)

This comprised 20 items of frequency of self-rated everyday memory failures (test–retest r = 0.86), with a score range of 0–80; higher scores represent less memory disability.

Knowledge and use of memory strategies: performance and self-report

Strategy Knowledge Repertoire (based on Troyer et al16)

This assesses the memory strategies familiar to the participant by a series of nine hypothetical memory situations requiring the application of a memory strategy (eg, remembering to pass on a message). Participants list as many strategies that could be used in each situation, and each strategy is scored by effectiveness and specificity: 2 points for each specific strategy appropriate to the memory situation, 1 point for a non-specific strategy and 0 point for an ineffective or non-strategy (inter-rater reliability r = 0.80 to 0.98). Higher scores represent more knowledge of effective strategy use. The score range is 0+; higher scores represent knowledge of more effective strategies. This task was undertaken by both the participant with aMCI and a family member.

Multifactorial Metamemory Questionnaire—Strategy subscale (MMQ-Strategy42)

Nineteen common memory strategies are self-rated for frequency of use (test–retest r = 0.88). The score range is 0–76; higher scores represent more frequent use of strategies.

Psychological well-being

Multifactorial Memory Questionnaire—Contentment subscale (MMQ-Contentment42)

Eighteen items describe emotions and perceptions linked to memory performance (test–retest r = 0.93). The score range is 0–72; higher scores represent greater contentment with memory.

Statistical method

A repeated-measures analysis of covariance (ANCOVA) model was used to evaluate the effects of intervention on data available at each time point for each outcome measure (prospective memory, strategy performance and psychological status). The between-subjects factor was group (intervention, waitlist), and the within-subjects factor was assessment phase (2 week and 4 month follow-up). Baseline scores and age were used as covariates to control for any baseline group differences in outcome measures and for age to reduce error variance. Baseline, 2-week and 4-month follow-up measures of each outcome measure were found to be approximately normally distributed; their standardised indices (z) of skewness and kurtosis for each group did not exceed ±1.96. Given the small sample size, an error rate (α) of 0.05 was adopted, without adjustment, for all tests of significant change between groups. Effect size estimates (Cohen d43) for the extent of group differences were tabled with descriptive statistics (M, SD) for each outcome measure, where values of 0.20, 0.50 and 0.80 correspond approximately to small, medium and large effects. In the reporting of ANCOVA results, effect sizes were expressed as η2 where 0.01, 0.09 and 0.25 correspond approximately to small, medium and large effects.43 All analyses were based on groups as randomised according to the intention-to-treat principle.44


Group characteristics

From a referral pool of 89 patients, we recruited 54 people with aMCI. A CONSORT flow diagram44 of the progress of the sample through the phases of the study and reasons for exclusions is provided in fig 1. By the final follow-up, there were 44 participants (ie, 82%) remaining from the initial sample. One participant with aMCI was unable to complete the intervention due to family ill health but still provided data and was included in “intention to treat” analysis.

Figure 1

CONSORT flow diagram of progression of participants through the study. “Withdrew” represents poor health, lack of time or lack of interest to participate. 1Participant did not complete full attendance at the intervention but did provide follow-up assessment data and was, therefore, included in the analysis.

The intervention (10 males, 12 females) and waitlist (nine males, 13 females) groups were broadly similar in distribution of sex, years of education, predicted IQ (Wechsler Test of Adult Reading—standard score33), general cognition (MMSE31) and Hopkins Verbal Learning Test—Revised28 34 (HVLT-R—total recall and delayed recall) (see table 1). The mean age of the intervention group was a few years older than the waitlist group, but this was accounted for, as age was a covariate in all analyses. In terms of accompanying family, there were 24 (mean age = 67.00, SD = 6.33) in the intervention group and 14 (mean age = 62.73, SD = 15.59) in the waitlist group.

Table 1 Demographic data on participant groups

Memory in everyday activities

Memory (dis)ability was evaluated through both performance of prospective memory tasks (prospective memory index) and a self-report questionnaire (MMQ-Ability) of everyday memory performance.

In the ANCOVA on prospective memory, the interaction effect was not significant, F(1,36) = 0.04, p = 0.835, η2 = 0.00; nor was the assessment phase effect, F(1,36) = 0.20, p = 0.656, η2 = 0.01. However, there was a significant medium-size group effect, F(1,36) = 5.98, p = 0.020, η2 = 0.14. From inspection of the means in table 2, the intervention group was performing significantly better than the control group on the prospective memory index at both 2 weeks’ and 4 months’ follow-up, even after accounting for any contribution of group differences in age or baseline performance.

Table 2 Memory in everyday activities

In the ANCOVA on self-reported memory ability, there was no significant interaction effect, F(1,40) = 0.00, p = 0.952, η2 = 0.00, group effect, F(1,40) = 0.35, p = 0.558, η2 = 0.01, or assessment phase effect, F(1,40) = 0.01, p = 0.941, η2 = 0.00.

Knowledge and use of memory strategies

Knowledge of memory strategies were evaluated for the aMCI participant and also for the family; use of memory strategies in everyday life were evaluated only from the self-report of the aMCI participant (see table 3).

Table 3 Knowledge and use of memory strategies

In the ANCOVA on self-reported strategy knowledge, the interaction effect was not significant, F(1,27) = 0.96, p = 0.336, η2 = 0.03; nor was the assessment phase effect, F(1,27) = 1.69, p = 0.204, η2 = 0.06. However, there was a significant medium-size group effect, F(1,27) = 4.33, p = 0.047, η2 = 0.14. Inspection of the means in table 3 indicated that the intervention group was performing significantly better than the waitlist group on strategy knowledge at both 2 weeks’ and 4 months’ follow-up, even after accounting for any contribution of group differences in age or baseline performance.

There was attrition in the number of family participants responding to the questionnaire on strategy knowledge at 4 months’ follow-up, so we restricted analysis to the baseline-2-week follow-up responses (see table 3). Using between-subjects ANCOVA, we compared family groups at 2 weeks’ follow-up on strategy knowledge, with age and baseline strategy knowledge as covariates. There was a large significant group effect, F(1,14) = 10.50, p = 0.006, η2 = 0.43. Inspection of the means indicated that family participants in the intervention group were performing significantly better than those in the waitlist group on strategy knowledge at 2 weeks’ follow-up, even after accounting for any contribution of family group differences in age or baseline performance.

In relation to the ANCOVA on self-report of memory strategy use in everyday activities, there was a significant group-by-assessment phase interaction, F(1, 40) = 4.21, p = 0.047, η2 = 0.10. In simple main effect analyses of group differences at each assessment phase, the significant difference between the Intervention group compared with the Waitlist group at 2 week follow-up, F(1,40) = 5.49, p = 0.024, η2 = 0.12, dissipated at 4 months’ follow-up, F(1,40) = 0.10, p = 0.749, η2 = 0.00. In a simple main effects analysis of assessment differences for each group, the Waitlist group showed no significant difference between 2 weeks’ and 4 months’ follow-up, F(1,40) = 0.44, p = 0.510, η2 = 0.01, whereas self-reported memory use in the Intervention group was significantly lower at 4 months’ than 2 weeks’ follow-up, F(1,40) = 5.42, p = 0.025, η2 = 0.12.

Psychological well-being

A self-report questionnaire (MMQ contentment) indexed emotions and psychological distress/contentment linked to memory performance in everyday life (see table 4).

Table 4 Self-reported MMQ contentment*

In the ANCOVA on MMQ-Contentment, there was a significant group-by-assessment phase interaction, F(1,40) = 8.46, p = 0.006, η2 = 0.18. In the simple main effect analyses at each assessment phase, group differences were not significant at 2 weeks’ follow-up, F(1,40) = 5.91, p = 0.09, η2 = 0.07 and at 4 months’ follow-up, F(1,40) = 0.65, p = 0.425, η2 = 0.02. In the simple main effects analysis of assessment differences for each group, the Intervention group showed no significant change from 2 weeks’ to 4 months’ follow-up, F(1,40) = 1.87, p = 0.179, η2 = 0.05, whereas self-reported contentment in the Waitlist group increased significantly over this period, F(1,40) = 8.23, p = 0.007, η2 = 0.17.


Overall our study results were encouraging for the value of early intervention for aMCI. Participants who received intervention demonstrated significant gains in actual performance of everyday memory as indexed by prospective memory tasks, that is, their memory disability reduced.

Effective prospective memory in everyday life depends on an interaction of planning and attention skills (working memory) and timely retrieval of previously learnt information (episodic long-term memory).45 This provides the basis for multiple opportunities for compensation of the impaired memory system in aMCI by the use of cognitive strategies that facilitate and strengthen encoding and retrieval of information. However, interestingly, this improvement in objective performance on prospective memory tasks was not replicated in a self-report of memory failures in everyday life, which remained relatively unchanged across assessments. It could be argued that although participants can improve their performance on our experimenter-induced prospective memory tasks, this did not generalise to self-generated tasks. However, patient testimony would not support this interpretation. For example, one participant described having learnt through application of strategies how to remember to self-administer her diabetic regime reliably so that she no longer required home nursing involvement; another participant reported his success in developing a mnemonic for remembering the rotation schedule for summer watering of sections of his garden. An alternative explanation is provided by a substantial body of research which reports little relationship between self-appraisal and objective memory performance, and indeed the self-appraisals were within the normative range.42 Explanations for this disjunction include the possibility that questionnaires are liable to be confounded by subjective memory beliefs or negative stereotypes about age-related memory.22 42 This suggests that there can be a time lag between changing behaviour (actual performance in everyday situations) and capacity for accurate self-appraisal of that change. A more detailed approach to self-report of day-to-day memory performance, a daily memory log for example, could improve accuracy of self-monitoring of change.

As compared with waitlist controls, participants had more knowledge and use of memory strategies following intervention at 2 weeks’ follow-up and fell within normative levels.42 Nevertheless, self-report of increased strategy use in everyday activities had dissipated by 4 months’ follow-up suggesting a need to investigate the use of ongoing booster sessions as in cognitive interventions for healthy older adults.17 Furthermore, the demonstration of sustained increased knowledge of effective memory strategies provides an enhanced basis for ongoing cognitive intervention.22

There were two secondary outcomes that provided additional support for the value of early intervention. First, family participants also demonstrated more knowledge about memory strategies following intervention. This is important, as there is emergent research about the significance of family management skills in relation to the behaviour of a family member with dementia.26 If many people with aMCI progress to AD, early mastery of skills in memory management by the family may represent an important component of optimising continuing family functioning. Determination of the ongoing significance of the impact of early cognitive intervention on family functioning will require a more extensive follow-up; however, the initial response by our family participants was encouraging. The results in relation to the other secondary outcome measure, memory contentment, require more caution in ascribing a positive intervention outcome. Although there was a strong trend for the participants with aMCI following intervention to self-report being more content with their memory at 2 weeks’ follow-up as compared with controls, this effect was not significant. However, almost universally, participants commented on how reassured they were that there were others with similar problems, and many continued to meet on an informal basis. To sustain greater self-confidence, as suggested for strategy use, additional booster sessions may provide an opportunity for re-enforcement of positive self-appraisal of the gains made in memory management in everyday life (cf, the findings of Winocur et al46). Nevertheless, it should be noted that the control group, as distinct from the intervention group, demonstrated an increase in contentment with memory between the 2-week and 4-month follow-up. This unexpected result may have reflected the fact that our informed consent procedures required full documentation of our experimental design, and the waitlist control group were anticipating joining an intervention at least by several months after participating in the study.

At least three methodological issues need to be considered in relation to interpreting the present results. First, replication with an increased sample size will ensure more reliable results and capacity to detect effects smaller than a medium size, which can be instructive in the developing field of cognitive intervention. It will also allow us to investigate individual differences in response to intervention. Clinically, this issue is important, as it is unlikely that this form of intervention will necessarily be appropriate for every person with aMCI. A second issue is that it will be helpful in further replications to consider using a social support group (but without focused memory intervention) as a control group to avoid the expectancies associated with a waitlist group as reported in our study findings, and also to determine the specificity of the memory rehabilitation in addition to any non-specific effects of therapeutic attention per se. However, it should be noted that in this preliminary study, we did select outcome measures that were specific to the focus of our cognitive training in order to target the efficacy of our memory rehabilitation programme. An additional limitation in relation to the design of our study is whether the research assessor remained masked to the participant assignment throughout the study. Participants were instructed not to discuss their study involvement with the research assessor, but the repeated follow-up assessment had potential to challenge the masking of assessment and introduce assessor bias. However, group differences were found on the primary outcome measure, prospective memory, which was an objective memory performance measure, and so we consider that this design approach does not invalidate our results. Finally, a third issue is that assessment of everyday memory disability through prospective memory is challenging. A central problem in neuropsychological assessment relates to establishing appropriate tasks that continue to be ecologically valid and yet reliably assessable.47 Our central prospective memory task index was limited to a scale of five points (0–4) which was not ideal and potentially blunted the sensitivity and reliability of the index, although it should be noted that our obtained effect of group differences on this index was moderately large. Performance of experimenter-led tasks within an office environment introduces some artifice, and continuing research will benefit from development of techniques for assessment of the functional impact of memory impairment in everyday activities, that is, memory disability.

In conclusion, in relation to the memory difficulties experienced in aMCI, families are seeking interventions that maintain independence and improve quality of life. Cognitive rehabilitation is a low-cost intervention that has been demonstrated in these preliminary findings to offer some moderation of the impact of failing memory in everyday activities.


The authors wish to acknowledge the support of the clinical staff in the memory clinics at Bundoora Extended Care Centre and Caulfield Hospital, especially M Winbolt and L Heeney.


View Abstract


  • Competing interests: None.

  • Funding: The research was supported by grants from Alzheimer’s Australia Research, Caulfield Hospital and La Trobe University.

  • Ethics approval: Ethics approval was provided by La Trobe University and participating hospitals.

  • Patient consent: Obtained.