Forgetting Rates of Prose Memory in Mild Cognitive Impairment

INTRODUCTION

The extent to which the slopes of forgetting in Alzheimer’s disease (AD) and in mild cognitive impairment (MCI) differ or not from healthy people is still a matter of debate. Previous research involving AD or MCI cohorts has provided contrasting findings, both in support [1–4] or against [5–9] the presence of faster forgetting in this clinical population. Divergent findings have been accounted for in term of methodological inconsistencies across studies [10] and to the variety of the testing methods [11].

However, contrasting findings are reported even within the same procedure. Here we consider prose memory. Walsh et al. [3] observed a greater decline in free prose recall among MCI patients within 30 min and 1-week delays, after equating their learning performance with that of the healthy controls (HC). In another recent cross-sectional study involving people with pre-symptomatic autosomal dominant AD, Weston and colleagues [4] reported faster forgetting on free recall of both verbal (wordlist and story) and visual (figures) material after 1-week, as compared to immediate and 30 min recall. Similarly, Zimmermann and Butler [12] observed faster forgetting on free recall of a wordlist after 1-week in a cohort of asymptomatic people at sporadic genetic risk (i.e., not inherited) to develop AD. On the other hand, Stamate et al. [9] employed a 70% correct encoding criterion on a task assessing cued recall for prose material and observed similar forgetting rates to controls, concluding that memory impairment in AD originates from an encoding deficit.

In the current study, we aimed to assess whether or not the forgetting slopes in a group of people with MCI due to AD (MCI-AD) are different from age-matched control participants by using a prose paradigm [13]. The term MCI due to AD refers to the symptomatic predementia phase of AD. In this phase of AD, people might experience a gradually progressive cognitive decline resulting from the build-up of AD pathology in the brain. Therefore, Albert et al. [14] considered this phase of impairment as a continuum, and they outlined the core clinical and research criteria for MCI-AD (see also [15, 16]).

METHODS

Procedure

All the participants were assessed through telephone-based interviews to avoid non-essential face to face contact. Prior to the first testing session, all participants and their respective caregivers were contacted by telephone to verify their interest and willingness in taking part to the study. All sessions were recorded. A flow chart of the experimental procedure is shown in Fig. 1.

OPEN IN VIEWER

Fig.1

Steps of the research paradigm comprising three sessions.

The experimental procedure was the same for both groups, with the only exception that people with MCI-AD were not administered any cognitive screen over the phone, as they had been already tested in the clinic on a previous face to face appointment, and that control group were tested with only one repetition of the story to obtain an average performance at immediate recall that was closer to that of the MCI-ADgroup.

On the first telephone-based session, the testing protocol was explained to the participants who were also informed about follow-up calls, without offering any information regarding the purpose of these phone calls. In the first session, people with MCI-AD were repeated the prose passage twice (one after the other) so to decrease the expected difference of performance at encoding between the two groups and to avoid floor performance, whereas HCs were read the story only once. The decision to present the MCI-AD group with two repetitions of the story derived from a pilot study where we compared the performance of four people with MCI-AD with one presentation vs two repetitions. During this pilot, we observed that, after two repetitions, MCI-AD participants were less likely to perform at floor at immediate recall and their performance was closer to the HC group.

Participants were instructed to listen carefully to the story and were told not to write any notes during and after its presentation. Immediately after, participants were asked to freely recall all they could remember about the story, without any specific instruction regarding specific events and without time limit. No interpolated task was used between the story presentation(s) and immediate recall.

In both follow-up sessions (after 1 h and 24 h), participants were contacted again by phone and asked to freely recall the story, without offering any cues or prompts. At the end of each follow-up testing session, participants were asked whether they tried to recall or thought about the story between the first and the second session, or between the second and the third session. This procedure was meant to check the potential role of active rehearsal. Overall, 35 participants reported having tried to recall or thought about the story among intervals at least once, 18 in the control group and 17 in the MCI-AD group. No participant reported writing down the story. At the end of the second follow-up session, participants were also debriefed about the experiment.

The performance of 4 participants from the MCI-AD group were excluded due to floor performance at immediate recall.

RESULTS

As revealed by one-tailed independent t-tests, the MCI-AD and the HC groups were matched for age, t(53) = 1.10, p = 0.13, and for level of education, t(53) = –0.28, p = 0.38.

Results and descriptive data are reported in Fig. 2 and Table 3 respectively. Table 4 also reports the odds ratios and the confidences intervals from the output of the generalized linear mixed model.

OPEN IN VIEWER

Fig. 2

Mean scores in Healthy Controls (HC, n = 26) and in the MCI-AD (n = 29) group at immediate recall, 1 h, and 24 h with Standard Errors.

As illustrated in Fig. 2, at immediate recall the HC and MCI-AD groups were not equated in terms of initial level of performance. Also, while HC group’s performance remained relatively stable across the three-time intervals, the MCI-AD group showed a steeper decrease in performance, especially from immediate to 1 h recall.

The between-subjects factor of group significantly predicted performance at immediate recall, as memory scores of the MCI-AD group were significantly lower when compared to the HC, b = –1.00, SE = 0.38, z = –2.63, p < 0.01, d = 0.55. This means that, despite being presented with two repetitions of the story, the MCI-AD group’s encoding was impaired at immediate recall.

Among the healthy controls, memory performance did not significantly decrease after 1 h, b = 0.03, SE = 0.26, z = 0.13, p = 0.89, d = 0.01, and after 24 h, b = –0.30, SE = 0.25, z = –1.18, p = 0.23, d = 0.16.

When considering the interaction between group and recall, memory performance of the MCI-AD group decreased significantly more than healthy controls from immediate to 1 h, b = –0.78, SE = 0.34, z = –2.26, p = 0.02, d = 0.43, and from immediate to 24 h, b = –0.91, SE = 0.34, z = –2.63, p < 0.01, d = 0.50.

To better explore these interactions, post hoc comparisons between 1 h and 24 h with a Bonferroni correction were carried out as implemented by the package emmeans [36]. Such comparisons revealed that memory performance of the MCI-AD group did not decrease significantly more than HCs from 1 h to 24 h, b = –0.12, SE = 0.34, z = –0.36, p = 1.00, d = 0.06.

Rate of floor performance (i.e., scores of 0) was prevalent in the MCI-AD group, which accounted for 13.8% of the scores after 1 h and 27.6% of the scores after 24 h (8 participants in total). After excluding floor performance, the MCI-AD group was still impaired at immediate recall, b = –0.81, SE = 0.36, z = –2.23, p = 0.05, d = 0.44. The overall shape of the forgetting rate slopes did not change, however the decreased statistical power resulted in differences short of significance.

As the nature of our memory task might have facilitated the performance of the HC group, the rate of ceiling performance (i.e., scores of 7) in HC was also evaluated. Scores at ceiling were 19.2% at immediate recall and after 1 h (5 participants), and 11.5% after 24 h (3 participants).

In the MCI-AD group, 20 participants reported items not presented in the original narratives (i.e., false memories), for a total of 59 instances, 12 at immediate recall, 24 at 1 h, and 23 after 24 h. Eight participants from the HC group also committed false memories, for a total of 15 instances, 7 at 1 h, and 8 after 24 h.

DISCUSSION

In this experiment, we aimed at investigating if people with MCI-AD present with different slopes of forgetting as compared to HC when assessed on prose memory via free recall and repeated testing.

We therefore compared the forgetting rates of a group of people with MCI-AD and a group of HCs matched for age and years of education. The MCI-AD group was presented with two repetitions of the story, whereas the HCs were presented with only one repetition. This procedure was devised to minimize floor scores at encoding in the MCI-AD group and to equate as much as possible their performance to the HC. Participants from both groups were then asked to freely recall the story immediately following encoding, and then 1 h and 24 h later.

People with MCI-AD showed poorer encoding at immediate recall and steeper forgetting of gist memory from immediate recall up to 1 h as assessed by free recall and with repeated testing of the material. However, forgetting between groups did not significantly differ from 1 h to 24 h. Therefore, the different rates of forgetting observed in the MCI-AD after 1 h could be due to a post-encoding deficit at immediate recall.

Encoding deficits in this population have been already documented in previous research [5, 37–39]. Furthermore, Moulin et al. [40] reported both encoding deficits and faster forgetting in people with amnesic MCI and AD despite showing practice effects during learning, yet these two deficits were not correlated with one another, leading the authors to postulate that they originated from two different mechanisms. In another study, Vallet et al. [7] observed that people with amnestic MCI and AD presented with an encoding deficit combined with a greater reliance on gist memory (see also [41]). Other studies showed normal forgetting on recognition memory but steeper forgetting rates on recall memory for lists of words [42–44] and prose passages [45, 46] in people with amnesia or AD. Green and Kopelman [44] observed that, when tested on free recall of words, amnesic patients presented with a primary encoding deficit and also with a deficit in long-term retention.

To account for these findings, it could be argued that, in our paradigm, people with MCI-AD appeared to benefit less than the HCs from repeated testing. Lack of practice effects after repeated testing has been reported with prose recall in this clinical population [47–52]. Lower practice effects on episodic memory in people with MCI are associated with an increased risk of progression to AD [53] and they also predict cognitive decline at one year follow-up [54]. Indeed, lower practice effects on episodic memory tasks reliably predicted future cognitive decline [55–58] and they were commonly reported in groups who tested positive to specific biomarkers, such as amyloid-β (Aβ) [59–61] and APOE ɛ4 [62]. Moreover, they have been observed within the same testing session [59], after one week [60], or up to more than a year [63, 64]. In support of the hypothesis of a differential benefit of practice effect in MCI-AD, in a recent review of the literature, Jutten et al. [48] observed greater average practice effects among healthy older adults compared to people with MCI [59, 65–67]. Nevertheless, given the absence of a direct comparison with and without repeated retrieval (i.e., unrepeated testing at immediate recall and at 1-day only, see [9]), this interpretation can only be based on a comparison between our data and previous literature.

However, a differential benefit from practice effects is not the only possible explanation of our findings. If people with MCI-AD benefitted less from retrieval practice than HC, we might have observed further differential forgetting between 1 h and 24 h. Between those time intervals there was more retrieval practice, as participants from both groups recalled the story for the second and the third time respectively. Yet, our results showed that, after an initial difference in the forgetting rates up to 1 h, people with MCI-AD did not show faster forgetting from 1 h to 24 h.

Therefore, the difference observed between the two groups could be due to a working memory impairment of people with MCI-AD. Previous studies included an interpolated (or filler) task between story presentation at learning and immediate recall [5, 69]. Given the absence of an interpolated task in our research paradigm, we are unable to control for working memory differences between groups. However, the use of an interpolated task between the stimuli presentation and immediate recall in a telephone-based paradigm would have likely confused our participants.

The current experiment presents with some limitations. Firstly, delayed memory scores could have been confounded by floor performance in some people with MCI-AD, especially after 24 h. However, if anything floor effect would act against the steepness of the forgetting slope. To prevent the issue of floor effect in the MCI-AD group, these participants could have been offered more repetitions at encoding. Nonetheless, this procedure would have biased our data due to the higher number of repetitions offered to this group as compared to HCs. Moreover, a score of zero has been chosen as floor. It should be noted that no participants from either group scored 1 at immediate recall.

Another confound might derive from ceiling scores among HC, especially at immediate recall. However, the number of healthy controls scoring at ceiling (i.e., 7 items out of 7) was small. According to Garin [70], floor and ceiling performance is considered problematic when more than 20% of participants score the minimum or the maximum score possible. In our experiment, the rate of HCs scoring at ceiling was never over 20% (equal to 5 participants).

Another caveat derives from the modality of presentation of the study material, as both the encoding and the retrieval sessions could only take place over the phone. Despite this being a validated methodology [9, 71–73], phone-based assessment might have affected the memory performance of our participants, regardless their cognitive status. This modality of presentation might have placed additional cognitive constraints on attentional control [74].

Lastly, in our experiment the initial level of performance between the two groups at encoding was not equated. Even though the MCI-AD group was offered a second repetition of the story, this could have not been enough for some participants whose performance was just above floor. On the contrary, previous studies in favor [3] or against [9] faster forgetting in early AD managed to equate the initial level of performance (see also [4, 46]).

In conclusion, when people with MCI-AD were tested on a prose recall paradigm requiring repeated testing, their encoding was found to be impaired at immediate recall and their memory for prose was also forgotten at a steeper rate after 1 h but not after 24 h. We argued that the differences observed between groups could be due to a post-encoding deficit, either caused by a differential benefit to repeated retrieval or by working memory differences at encoding.