Participants

A total of 49 older adults (27 women and 22 men; age = 71 to 87 years; M = 78.7; SD = 3.9) participated in the research. Participants' responses on a self-report questionnaire (form available from the corresponding author) determined their eligibility for inclusion in the study.

All selected participants were retired middle class healthy community-dwellers who lived independently and functioned successfully. All had been successful in their professions, or positions in the household. At the time of the study, participants expressed subjective complaints of memory or cognitive dysfunction. However, neither they nor their treating family physicians had considered referral for assessment of dementia. Their subjective problems did not affect their daily functioning.

Their level of normal functioning was confirmed at preassessment by their Mini-Mental State Examination (MMSE) score. Despite their subjective complaints, the participants did not fit the criteria for mild cognitive impairment (Petersen et al., 1999). When an estimate of general intelligence (NART-R) was assessed in relation to memory (Logical Memory), only 2 participants had memory scores that were greater that 1.5 SD different from their IQ scores, and the memory scores of these 2 persons were less than 1.5 SD different from normative data. An additional criterion was fluency in written and spoken English. Our formal assessments confirmed the participants' self-reports of general good health commensurate for their age. There was no history of significant medical, psychiatric, or neurological problems; excessive alcohol use or use of nonprescription drugs; and current use of central nervous system–altering medications.

All participants were competent, were presented with an approved information form, and gave their signature consent for participation. Participants were volunteers from the community at large who were recruited through advertising and word of mouth. Recruitment took place over time due to the practical necessity of administering the protocol to several groups. The volunteers who met inclusion criteria were divided into an Early Training Group (ETG; N = 29) and a Late Training Group (LTG; N = 20). Random assignment to the two groups was qualified by the need to equalize the groups with respect to MMSE (all 25 or higher, groups matched ± 2 points), education (approximately equal numbers within the following grade ranges: grade 8 or lower; 9–10; 11–12; college or higher), sex, and age (all were within the age range of 72–87 years, groups matched ± 3 years). Thus a blocked randomized procedure was used to maintain an equal balance in both groups on these factors. The ETG entered the rehabilitation program shortly after screening; the LTG first acted as a control group for 3 months, and then underwent the protocol (hence “Late” Training Group). In the control procedure, contact was maintained with participants through an individual meeting with the group leader and by telephone. In addition, participants underwent regular testing according to the prescribed schedule. Missed attendance was negligible. On the few occasions that it did occur, participants were able to make up the session by attending another group or through individual contact with the group leader, with the result that they received the full training protocol, including the one-on-one sessions, with exceptions outlined in Figure 1.

CONSORT-type figure illustrating the maximum number of participants at each assessment. The loss of subjects in the Late Training Group (LTG) for Assessments C and D was secondary to a province-wide pandemic (SARS) which closed hospitals and associated buildings for all nonessential personnel (such as research participants), as indicated in the text. ETG = Early Training Group.

Before admission to the program, volunteers were administered the following neuropsychological tests (average time of administration = 1.5 to 2.0 hr) to characterize various aspects of the groups' cognitive function (baseline testing) and provide a screen for cognitive impairment and general psychological status:

1. Mini-Mental State Examination (MMSE; Folstein et al., 1975): a brief assessment of general functioning that screens for dementia. A score greater than or equal to 24 was required for inclusion;
2. National Adult Reading Test–Revised (NART-R; Nelson & Willison, 1991): a reading test that provides an estimate of intellectual status. A minimal score of 90 was required for inclusion;
3. Digit Span (forward and backward; Wechsler, 1987), forward span measures simple attention, with backward span providing a general index of working memory;
4. Logical Memory (Wechsler, 1987), immediate recall of two short stories;
5. Judgment of Line Orientation (JLO; Benton et al., 1978), a measure of visual spatial judgment considered primarily sensitive to right parietal lobe functioning;
6. Wisconsin Card Sorting Test (WCST; Heaton, 1981; administration following the procedure of Stuss et al., 2000b), a measure of concept learning and switching of concepts following feedback. This test is widely held to be sensitive to frontal lobe functioning if comprehension and visual attention are normal (Stuss et al., 2000b). Our version, derived from the original instructions as described by Grant and Berg (1948; see also Milner, 1963) is similar to the standard version as implemented by Heaton (1981), with the exception that all 128 cards are administered, providing a more sensitive measure of loss of set;
7. Boston Naming Test (BNT; Kaplan et al., 1983), a visual confrontation task of naming of objects;
8. Beck Anxiety Index (BAI; Beck & Steer, 1990), a self-report index of anxiety symptomatology;
9. Geriatric Depression Scale (GDS; Brink et al., 1982), a test of depression commonly used in the elderly population.

Means and standard deviations for ETG and LTG for age, years of education, NART-R, and MMSE are listed in Table 1. The general neuropsychological measures served several purposes as an additional screening beyond the self-report of the participants: comparison of equality of groups; comparison of participants to normative data; and additional verification of normal level of functioning (data were examined by a board certified neuropsychologist if there was question of impaired functioning; no subject was rejected on the basis of the general neuropsychological measures). The two groups were comparable in virtually every aspect. There were no differences in age, education, sex, intelligence, or scores on the MMSE. The groups were also comparable in general functioning (no significant group differences between the ETG and LTG on the general neuropsychological battery), with the exception of a solitary significant difference in immediate (but not delayed) recall on Logical Memory Stories in favor of the ETG. This general group equivalence is an important prerequisite for interpreting any group differences secondary to the interventions. In addition, on none of the measures did either group differ more than 1.5 SD from the mean score of the published normative data (see Table 1). The lowest scores were in logical memory, but this finding was considered to reflect their memory complaints. In other words, this was a very representative sample of normal older adults living independently and functioning well in the community, who had noticed changes in their cognitive abilities and who wanted to do something to recover apparently diminished function.

Demographic, neuropsychological screening and normative data summary

The project was approved by the University of Toronto/Baycrest Research Ethics Committee, and followed the Canadian Tri-Council Ethics Guidelines as well as the Declaration of Helsinki on Ethical Principles for Medical Research Involving Human Subjects.

Experimental Design

A within-subject, crossover design was used (see Figure 2). To measure the efficacy of our rehabilitation program, after pre-admission testing, there were four major assessment sessions (each lasting approximately 2.5 to 3.0 hr) identified as Assessments A to D. The group leaders were not involved in the assessment sessions. The complete battery, consisting of experimental neuropsychological, practical task planning, and psychosocial tests, was administered before (Assessment A) and after (Assessment B) the ETG underwent the training protocol. Assessment A indicated whether the two groups were equivalent before training; Assessment B provided a test of the effects of rehabilitation training of the ETG, relative to the LTG whose members performed the control procedure.

Within-subject crossover experimental design with the timeline for the early and late training groups.

After Assessment B, there was a crossover in which the LTG switched from the control condition to rehabilitation training, and the ETG from training to the control condition. Following the completion of training for the LTG, the test battery was administered to both groups again (Assessment C).

Both groups received follow-up testing 6 months after completion of training (Assessment D) to assess long-term benefits of rehabilitation (see Baltes & Willis, 1982). The difference in delay (see Figure 1) between Assessments C and D for the ETG and LTG was necessary to ensure the same delay between the end of training and the long-term follow-up test.

The tests used in Assessments A through D were as follows: Cognitive/Memory: Alpha Span, Brown-Peterson, Hopkins Verbal Learning Test, Logical Stories Memory, Verbal and Semantic Fluency; Goal Management: simulated real-life tasks (organizing car pools, swimming lesson); Psychosocial: Dysexecutive Function Test, Everyday Activities, Locus of Control, Life Orientation Test, and the Memorial University of Newfoundland Scale of Happiness. The descriptions and references for these tests are provided in the corresponding papers. As explained in the specific papers, alternate forms were used, where feasible and appropriate, We decided in favor of understanding the effects of test order rather than counterbalancing because of the relatively limited sample size, common in studies of this type.

In addition, we administered a self-assessment questionnaire (SAQ) at the end of the training program querying how well participants had responded to rehabilitation training. The SAQ consisted of statements, of which 15 were related to the Psychosocial component (e.g., “I am leading a more active lifestyle”): 8 statements to the goal management component (e.g., “I am better at setting goals for myself”), and 7 to the Memory component (e.g., “My memory seems better because I am concentrating more”). Participants were asked to respond to each statement based on how they felt “today as compared to last month,” according to a 4-point scale. The maximum total score was 120, distributed across the three components as follows: Psychosocial, maximum = 60; modified GMT, maximum = 32; Memory, maximum = 28. The SAQ results are summarized in the Overview paper (Winocur et al., 2007a)

The Training Protocol

The program was administered to groups of 5–6 participants, with each group run by an experienced leader who had been trained specifically to lead the sessions. Before the start of the program, individual goals were identified for each participant to establish personal motivation and to ascertain reasons for his/her involvement (Prigatano, 1999). In addition, to ensure engagement and active participation, the group leader met for approximately 1 hr each time with each participant individually over the course of the program (ETG, five times: one before start of program to identify goals, one after the third session of each module, and one at the end of the program to evaluate progress related to goals; LTG, 6 times: because of the delay in starting training, an additional session was held before the start of their training program to see if they made progress on their goals on their own during the control wait period, and to reestablish goals for the actual program).

The entire rehabilitation program spanned 14 weeks and included an introductory seminar, 12 weekly training sessions, and a wrap-up seminar. Test sessions and individual meetings with participants were conducted during this period, as scheduled (see Figure 1). The training program consisted of three distinct modules, each consisting of 4 consecutive weeks. The groups met once per week in 3-hr sessions. The modules were presented in fixed order: Memory Training, modified GMT, and Psychosocial Training. The contents of each module are presented in the papers related to the respective modules. Each training session adhered to a prescribed format and was highly interactive, with the group leader assuming the roles of instructor and facilitator. Most sessions included practical projects related to the theme of the module.

To maximize learning and to generalize the training to home environments, participants were given home assignments of approximately 1 hr in duration, to complete during the week. These assignments were designed to encourage participants to apply strategies and techniques learned in that week's training session to their day-to-day lives. To ensure completion, the home assignments were monitored for each participant in the weekly sessions.

Statistical Analyses

We compared baseline performance between the two groups using analysis of variance to test for possible differences on each of the measures at Assessment A. Because volunteers were assigned to the two groups using a block randomization procedure, we predicted that there would be no differences between the groups on any of the measures at baseline.

At each of the two subsequent assessments (B and C), we investigated differential change between the two groups on each measure relative to performance at the preceding assessment using analysis of covariance. Effect sizes (η²) are reported alongside these significance tests of differential change between groups. To indicate the size of the effect, following the definitions of Cohen (1988): small, η² = .01; medium, η² = .06; large, η² = .14. Hypothesis tests were set at an α-level of 5%. To avoid unacceptable loss of power in this exploratory study, adjustments for multiple comparisons were not made. Accordingly, however, appropriate caution was taken in interpreting the results. Where informative, we also present results of repeated measures analysis of variance within groups to provide a further index of the effect of rehabilitation on performance over consecutive paired assessments.

At Assessment B, we predicted a rehabilitation training effect in the ETG relative to the LTG, which acted as the control group at that point. Analysis of covariance performed on the dependent measures at Assessment B included Assessment A performance as a covariate to increase precision by accommodating natural heterogeneity between subjects within a group. It was predicted that within-group analyses would show improvement in the ETG but not the LTG.

At Assessment C, we expected a rehabilitation training effect in the LTG relative to the ETG. No change was expected in the ETG because they were anticipated to be in “maintenance” mode. Because Assessment C scores were adjusted to compensate for differences in Assessment B scores in this analysis of covariance, we predicted a significant difference between groups in favor of the LTG that had just received rehabilitation training.

Assessment D provided an examination of the long-term benefits of rehabilitation training for each group. To evaluate improvement between Assessments A and D, we performed either repeated measures analysis of variance on the two scores or linear regression of adjusted Assessment D score on adjusted Assessment A score. For the latter, with both measures adjusted by subtracting mean Assessment A score, the intercept for this regression model was interpreted as a relative change between assessments.

Some inability to obtain data at Assessments C and D was encountered for the last subset of LTG participants. This loss of data was the result of logistical problems and was not due to motivation of the volunteers. After the last LTG subgroup had received Assessment B, and after their training sessions but before Assessment C, severe acute respiratory syndrome (SARS) broke out in Toronto and, as a result, essentially all hospitals in Toronto were closed for several months. Research subjects were not permitted to enter the hospital or research buildings. Because this development precluded data collection for Assessments C and D for one of the LTG subgroups, we did not consider it to be related to the efficacy of the rehabilitation program. Nevertheless, the reduction in the number of participants does raise the issue of power, and the long-term follow-up data should be interpreted with caution. For these reasons, our analytic approach should be considered an observed cases approach, and not intent-to-treat. A degree of data reduction was achieved using composite scores where multiple measures were correlated and were hypothesized to be nonhierarchical (e.g., G7 composite score in the Psychosocial paper; Winocur et al., 2007b).

Predictions

The following general predictions guided the trial:

1. At Assessment A, there would be no differences between the groups on the defined measures.
2. At Assessment B, group differences favoring the ETG were expected. This finding would indicate that there was a specific effect of rehabilitation for the ETG because, at this point, the LTG would be acting as a control group.
3. At Assessment C, after the LTG had received training, there would be no difference in the two groups' overall performance. Statistically controlling for performance at the prior assessment, however, would allow analyses to show training-related group differences. It was expected that a within-group analysis from Assessment B to C would reveal improvement in the LTG, confirming the effect of training.
4. At Assessment D, there would be no group differences. While some decrease in scores was expected at Assessment D, relative to those recorded immediately after rehabilitation training, it was predicted that both groups would perform better at Assessment D, as compared to Assessment A, thereby reflecting long-term benefits of training.
5. Training would have direct benefits on frontal lobe-mediated strategic processes with derivative benefits to cognitive functions associated with other brain regions (e.g., episodic memory).