Abstract
The prognostic value of mild cognitive impairment (MCI) is being questioned, with some MCI subjects reverting to normal cognition (NC). The reversion rate varies mostly depending on the study design, the setting, and both MCI and NC definitions. Previous studies have focused on the profile of subjects who revert to NC, but the role of comorbidities has not been entirely investigated. We aimed to evaluate the proportion of MCI subjects who revert to NC in a memory clinic context, focusing on the role of comorbidities. Between 2004 and 2013, 374 MCI subjects were recruited. During a mean time of 32 ± 25.5 months, 21 subjects (5.6%) reverted to NC. Subjects who reverted to NC were younger (p = 0.0001), more educated (p = 0.0001), had a better global cognition (p = 0.0001), as assessed by the Mini-Mental State Examination (MMSE) and suffered from more comorbidities (p = 0.002), as assessed by Cumulative Illness Rating Scale (CIRS) than those who developed dementia. The Cox Regression Model, constructed to adjust for the confounders, showed that the higher were the MMSE (HR = 1.83, CI 95%: 1.07–3.11) and the CIRS score (HR = 1.3, CI 95% 0.88–1.92) at baseline, the higher was the probability of returning to NC than developing dementia, though the last association was not significant. Subjects who reverted to NC were more frequently affected by respiratory (p = 0.002), urologic (p = 0.012), and psychiatric (p = 0.012) diseases. The cognitive performance of subjects with medical comorbidities could benefit from preventive strategies aimed at treating the underlying diseases.
INTRODUCTION
Mild cognitive impairment (MCI) is considered either a risk factor or a prodromal phase of dementia. It is defined as a cognitive dysfunction more severe than expected for age and level of education, but not affecting main activities of daily living [1].
However, the prognostic value of MCI is currently being questioned, considering that many subjects with MCI may remain cognitively stable over time, and some do even revert to normal cognition (NC). Identifying those factors that in subjects with MCI predict the progression to dementia or the reversion to NC is essential to prevent or delay cognitive decline and to support, where possible, the regression to normality. Reported rates of reversion to NC may widely vary across studies based on the source of the population considered, and due to several other aspects.
Such variability in reported reversion rates mainly depends on the study design and setting, with higher rates, ranging from 29% up to 55% , reported in population studies [2 –5] and lower rates, ranging from 4% to 15% , reported in clinical studies [6 –10]. The variability in reported outcomes can also be due to an equally high heterogeneity in the definition of cognitive impairment [11]. The “memory clinic approach”, for example, aims at maximizing the accuracy in identifying those subjects with cognitive decline who will develop dementia within the next 2 to 3 years (Questionable Dementia (QD) [12]; Age Related Cognitive Decline (ARCD) DSM IV [13]; MCI [1]). The different diagnostic criteria for cognitive impairment are summarized in Table 1.
Results from the literature show that different definitions of cognitive impairment lead to different estimates of the prevalence of the considered condition, and to different rates of reversion to NC.
The inclusion and exclusion criteria applied, such as presence of depressed mood or reversible medical causes (as the presence of depression, metabolic derangement, cerebrovascular disease) [14], and the length of follow-up considered (ranging from 1 to 4 years) are further aspects increasing the variability of the reported reversion rates.
Moreover, the definition of “reversion to normality” is not uniform across studies, nor clearly defined. Reversion to NC should be identified based on a defined set of neuropsychological criteria, but, despite the need of such codification, no agreement still exists on the number and type of tests to be used. NC can currently be defined using an extensive neuropsychological battery [6 , 15], or simply based on the score of the Mini-Mental State Examination (MMSE) [10].
Thus, defining the specific characteristics of the subjects with MCI who will revert to NC is essential, in this context, to identify possible risk factors affecting cognition that could be prevented.
Previous studies have usually taken into consideration medical and/or psychiatric causes when defining inclusion and exclusion criteria for the enrolment of subjects with MCI, while paying little attention to comorbidities. Therefore, in the present study, we aimed at assessing the proportion of subjects with MCI who would revert to NC from the perspective of our memory clinic focusing on the role of comorbidities. We hypothesized that MCI subjects who suffered from more comorbidities may have a higher probability of reverting to a state of NC rather developing dementia.
MATERIALS AND METHODS
Study population
This prospective cohort study was carried out in the Center for Research and Treatment of Cognitive Dysfunctions of the Luigi Sacco Hospital, University of Milan. The study was approved by the local Ethics Committee and all participants signed an informed consent.
A total of 3268 consecutive subjects were examined between September 2004 and May 2013 in our memory clinic, and 503 (15.4%) out of them met the diagnostic criteria for MCI [1] and were included in the study. Subjects were classified as MCI if they had subjective cognitive complaints preferably confirmed by an informant, objective cognitive impairment defined by a pathological score on at least one task of an extensive neuropsychological assessment (see “neuropsychological assessment” for details), unimpaired basic activities of daily living [16] and unimpaired or minimally impaired complex instrumental activities of daily living [17], unimpaired general cognitive functioning defined by a score of 0.5 of the Clinical Dementia Rating scale (CDR) [12], and absence of dementia according to the DSM IV criteria [13]. Subjects were then further classified as “amnestic MCI” if they had memory impairment alone, as “non-amnestic single domain MCI” if they had an impairment in one cognitive domain other than memory, and as “multiple domain MCI” if they had an impairment in at least two cognitive domains [1].
Subjects with a diagnosis of dementia according to the DSM IV criteria [13], any psychiatric disorder, severe cerebrovascular disease, significant history of head injury, any major systemic illness or medical complication including vitamin deficiencies, thyroid disorders and sensory disorder (i.e. blindness or deafness), a history of drug or alcohol addiction or abuse, and structural brain alterations including mass lesions and hydrocephalus were considered non-eligible for the present study.
Participants were assessed at baseline through an extensive evaluation. Baseline assessment included medical and neurological examination, laboratory testing, neuroimaging and neuropsychological assessment. Subjects were also offered APOE genotyping, and the results from the tests of those who accepted were included in baseline data.
Comorbidity assessment
Comorbidity was assessed using the modified Cumulative Illness Rating Scale (CIRS) comorbidity index [18]. The modified CIRS includes 14 categories assessing the impairment of each organ system, with a score ranging from 0 to 4:0 means that the system considered has no clinical dysfunction or has a clinically irrelevant dysfunction; 1 means that the system considered has a mild clinical dysfunction, or has had a relevant dysfunction for which no treatment is required; 2 means that the subject has a moderate clinical condition/disability that requires a first-line treatment; 3 means that the subject has a severe condition and/or disability and/or a chronic condition requiring a complex treatment; 4 means that the subject has a very severe condition and/or needs an urgent treatment due to an organ failure and/or severe functional disability. The total score (TSC) is calculated adding the scores from each of the 14 individual system scores. Thus, the total score ranges from 0 to 56, though too high scores are unlikely as they would mean multi-system failures, which are not compatible with life. The comorbidity index (CM) was defined, based on these scores and according to Salvi et al. [18], as the number of categories for which a score of 2 or above was reported.
Neuropsychological assessment and definition of NC
Subjects with MCI were assessed at baseline and at follow-up visits through an extensive battery of widely used neuropsychological tests, according to a standardized procedure. The battery was intended to assess the whole cognitive spectrum including memory, language abilities, visuo-spatial and visuo-perceptual abilities, executive functions and attention (Table 2). Global cognition was measured using the Mini Mental State Examination (MMSE) [19].
Memory tasks
the Story recall (SR) [20] and the Rey Auditory Verbal Learning Test (RAVLT) [21] were used as independent measures since both were considered able to represent the theoretical construct of episodic memory. The SR is a measure of verbal memory. The test require participants to listen to a brief story and provide immediate recall, then they are told the story a second time, engaged in different activities for 10 min, and then required to provide delayed recall. The RAVLT is structured in five subsequent trials during which participants are asked to learn a list of 15 unrelated words, and required after 15 min to provide delayed recall. The battery, along with the two tests for episodic memory, included also the Rey Complex Figure (RCF) recall [22], which is a measure for long-term visuo-spatial memory requiring participants to complete a task involving also constructional abilities.
Non-memory tasks
Letter Fluency [21] and Category Fluency [23], which require participants to generate lists of words, have proven to be sensitive indicators of lexical-phonological functions and cognitive flexibility. Category fluency is largely based on semantic processing, whereas attention and lexical knowledge have a more extensive role in letter fluency [24]. The Rey Complex Figure copy [22] test measures perceptual organization, constructional praxis and copy strategy. The Clock Drawing Test [25] requires different cognitive abilities, including visuo-spatial abilities, abstract conceptualization and executive control. The Raven Coloured Progressive Matrices 47 (CPM47) [21] tests the processes of hypothesis generation and abstract reasoning. The Trail Making Test (TMT) [26] assesses visual search speed, psychomotor speed and attention in part A, and task switching in part B. The Frontal Assessment Battery (FAB) [27] includes a set of tasks to assess conceptualization, item generation, motor sequencing, sensitivity to interference, inhibitory control and environmental autonomy. The Stroop Colour-Word Test [28] assesses cognitive flexibility, and the ability to inhibit automatic responses. Scoring is based on time and the number of errors.
Tests were administered in a standardized sequence, alternating verbal and non-verbal tests. The sequence is aimed at avoiding the risk of contamination in memory tests. Therefore, tests that include contents that could affect the performance on a memory test were not administered in the time span between an immediate recall and a delayed recall performance. All test scores, where appropriate, were adjusted for age and educational level according to available literature.
The definition of NC was based on neuropsychological criteria. Subjects were defined NC if they had an MMSE score adjusted for age and level of education [19] ≥24, and showed no impairment in all tasks included in the neuropsychological assessment, basing on the Equivalent Scores, a 5-point scale [29] in which scores lower than 1 are considered pathological.
Depression was assessed using the Geriatric Depression Scale (GDS) [30].
Follow up
Subjects were reassessed at 12-months intervals to monitor the course of MCI. Reassessments included both the clinical and the neuropsychological tests. Subjects were defined as “reverters” to NC when they scored normal in all the tasks included in the neuropsychological assessment, and resulted having normal functioning in both ADL and IADL. A diagnosis of dementia [13] required evidence of objective cognitive impairment limiting the independence in performing activities of daily living. Subjects diagnosed with dementia were classified according to standardized criteria, based on the type of dementia, as having AD [31], dementia with Lewy bodies [32], Parkinson’s disease dementia [33], frontotemporal dementia [34], and vascular dementia [35]. The length of follow up was expressed in person-months and calculated as the time interval from baseline assessment to the last annual follow up assessment for the subjects who remained cognitively stable or reverted to NC, and as the time interval from baseline assessment to the time of diagnosis for the subjects who progressed to dementia. November 2014 was set as the end of the cohort surveillance.
Statistical analysis
Continuous measures were reported as mean, standard deviation, range, median and interquartile range (IQR). The continuous measures were compared with the ANOVA test and post hoc analyses were made when appropriate.
Categorical variables were reported as absolute number and percentage. Categorical variables were compared with the Pearson chi squared test.
We investigated the association of baseline demographic, social, cognitive aspects and comorbidity with the rate of reversion to NC respect to dementia or stable MCI using Cox proportional-hazards regression models to estimate hazard ratios (HRs) with 95% confidence intervals (CIs). The Cox models included as covariates the variables statistically significant at univariate analysis, considered also significantly related to the outcome.
For the imputation of missing data we have assessed if there were systematic differences between missing values and the observed values (missing completely at random) [36]. Subsequently we have randomly attributed to the cohort of subjects with missing values the same variable frequency stated in the group with values observed.
All statistical analyses had an alpha value of 0.05 and were performed with SPSS [37].
RESULTS
Fifty-five out of 503 subjects included in the study were lost at follow up. Cognitive status was assessed administering CDR by phone in 74 subjects who were consequently not included in the statistical analyses. The proportion of males in this subset of subjects was higher (p = 0.013), as was the frequency of atrial fibrillation (p = 0.022) as compared to the group that completed follow-up. No other differences in any baseline characteristic, including the distribution of MCI subtypes, were observed between this subgroup and the remaining subjects who reached the end of the study.
Three hundred and seventy-four subjects were finally included in the study.
Subjects were followed for a mean of 32.0 ± 25.5 months (range 1.7–51.1; median 24.4, IQR 14.2–40.3 months). During this time, 21 subjects (5.6%) reverted to NC, 110 (29.4%) remained stable MCI, and 243 (65%) progressed to dementia. AD was diagnosed in 156 (64.2%) subjects, vascular dementia in 26 (10.7%), mixed dementia (neurodegenerative and vascular) in 40 (16.5%), dementia with Lewy bodies in 16 (6.6%), Parkinson’s disease dementia in 1 (0.4%), frontotemporal dementia in 2 (0.8%), and dementia due to multiple causes (neurodegenerative, vascular, toxic) in the remaining 2 (0.8%) patients. Fifty subjects were diagnosed with dementia within 12 months from the baseline, 84 between 12 and 24 months, and 109 after 24 months of follow up. The mean time to diagnosis of dementia was 27.1 ± 19.5 months (range 1.7–99; median 22.3, IQR 13.1–35.0).
Subjects who remained cognitively stable were followed up for a mean of 38.4 ± 29.8 months (range 5.6–151.1; median 30.6, IQR 15.9–30.6), while subjects who reverted to NC were followed for a mean of 58.0 ± 39.8 months (range: 12–130.1; median 42.3, IQR 23.8–98.0).
Results from the ANOVA (Table 3) showed statistically significant differences (p = 0.001) in mean age at diagnosis among the three groups (stable MCI, NC, dementia). At post hoc analyses the subjects who reverted to NC were younger than those with stable MCI (p = 0.046) and those who progressed to dementia (p = 0.001); subjects with stable MCI were also younger than those who progressed to dementia (p = 0.014).
The proportion of females in subjects who reverted to NC (28.6%) resulted lower when compared to subjects who remained cognitively stable (63.6%) and those who were diagnosed dementia at follow up (61.3%) (p = 0.009).
At post hoc analyses subjects who reverted to NC were more educated than those who remained stable MCI and those who progressed to dementia (p = 0.001 for both comparisons).
Subjects who reverted to NC had a better global cognition than the ones who remained cognitively stable (p = 0.002) and those who progressed to dementia at follow up (p = 0.001); subjects with stable MCI had a higher MMSE score than those who progressed to dementia (p = 0.001).
The frequency of multiple domain MCI resulted significantly higher in the group of subjects who were diagnosed with dementia at follow up when compared to the group of subjects who reverted to NC (68.7% versus 38.1%), while the frequency of the non-amnestic single domain MCI subtype resulted significantly higher in the group of subjects who reverted to NC when compared with the group who was diagnosed with dementia at follow up (33.3% versus 11.6%) (p = 0.004).
Statistically significant mean differences between groups were also observed in the CIRS CM (p = 0.002). Results from post hoc analyses showed that subjects who developed dementia had less comorbidities when compared to subjects with stable MCI (p = 0.002).
APOE genotype was available for 252 subjects (64.4%). Statistically significant differences in APOE ɛ4 genotypes between groups (p = 0.005) were observed. Specifically, the frequency of the ɛ4 allele resulted higher in the group of subjects who developed dementia at follow up (42.1%) than in the group with stable MCI (21.9%) and the group who reverted to NC (20%) (Table 3).
Proportional hazards condition of first Cox model show a chi-square of 19.1 (p = 0.014). Results from this Cox model (Table 4) show that higher MMSE scores at baseline are associated to a higher probability of reverting to NC at follow up than to develop dementia (HR = 1.83, 95% CI: 1.07–3.11).
This model included data from the subset of 179 subjects for which enough information for all the covariates, including APOE genotype, was available. The model shows a higher risk of reverting to NC for those with higher CIRS CM, though the association is not significant (HR 1.3, 95% CI 0.88–1.92) (Table 4). Thus a further simulated model was designed including data from a larger sample (260 subjects) assuming that the 81 subjects for which data on the APOE genotype were missing had the same distribution of ɛ4 genotype as the group for which data were available. The group of subjects with APOE genotype data missing did not show statistically significant difference in all baseline characteristics reported in Table 3 respect to the group of subjects with APOE genotype data observed. Results from this last model show that subjects with a higher CIRS CM at baseline had a statistically significant higher probability of reverting to NC (HR: 1.50; 95% CI: 1.03–2.18) (Supplementary Table 1).
Proportional hazards condition of second Cox model show a chi-square of 27.4 (p = 0.001). Results from this second Cox model (Table 5) show a higher probability of reverting to NC than remaining cognitively stable for subjects with higher MMSE scores (HR = 1.52, 95% CI 1.03–2.22) and a higher education (HR = 1.16; 95% CI 1.00–1.34). Results from the sub-items of the CIRS are reported in Table 6 and show a higher frequency of respiratory (p = 0.002), urologic (p = 0.012), and psychiatric conditions (p = 0.012) in the group of subjects that reverted to NC at follow up than in those subjects who developed dementia.
DISCUSSION
In a memory clinic context, it is crucial to identify the variables that can predict the reversion to NC in MCI subjects in order to allow the clinicians to focus on those subjects with specific characteristics that have higher probability to revert to NC. The present study demonstrates that 5.6% of subjects with MCI reverted to a state of NC. Our results are consistent with the few available prospective cohort memory clinic studies who used similar MCI [1] and NC reversion definition. In particular, Gallassi et al. [6] and Nordlund et al. [9] reported a 4% and 4.5% rate of reversion from MCI to NC, respectively.
The present study showed an association between reversion to NC and CIRS comorbidity score. Nevertheless, the univariate association between individual baseline CIRS items and cognitive status at follow-up does not allow us to infer that those items are the main determinants of the association found at the multivariate analysis.
In the multivariate analysis, for each increase of one point of MMSE reported at baseline, the probability of returning to NC was 83% at the follow-up; moreover, those subjects who had a higher CIRS CM at baseline had a 30% of probability for each increase of one point of reverting to NC rather than developing dementia in the follow-up. This latter observation, even though not statistically significant in the model showed, may be relevant in a clinical standpoint considering also the statistically significant differences reported in scores of single items of CIRS (Table 6). We have hypothesized that the association observed with CIRS CM might have not reached statistical significant due to the reduced sample size: The APOE genotype had been determined in 64.4% of the whole cohort, as the test was not included in the routine workup but offered on a voluntary basis. Thus, in the simulated model [36], for each increase of one point of CIRS CM reported to baseline the probability of returning to NC was 50% when compared to subject who developed dementia; MMSE, education, and non-amnestic single domain MCI subtype confirmed a protective role.
The respiratory, urologic, and psychiatric diseases were more frequently observed in subjects who reverted to NC than in the ones who developed dementia (Table 6). These diseases may play a role in the cognitive performance of the subjects. As a result, when the underlying condition is treated, a subsequent improvement in neuropsychological performance could be observed.
For what concerns respiratory diseases, obstructive sleep apnea syndrome (OSAS) and chronic obstructive pulmonary disease (COPD) have an active role in the development of cognitive impairment. The primary mechanisms implicated are nocturnal hypoxemia, sleep fragmentation, and daytime sleepiness [38, 39]. Large studies suggest that hypoxemia and hypercapnia observed in COPD and OSAS patients are responsible for frontal impairment and executive dysfunction and subsequent worsening in cognitive performance [40, 41].
The association between urologic disorders and cognitive performance is less clear and the available scientific literature lacks of evidence. In our cohort, all the subjects who suffered from urological disorders were male and were affected by prostatic diseases (mostly benign prostatic ipertrophy). The sleep fragmentation due to nycturia could affect cognition and influence the cognitive performance of such subjects [42].
Finally, psychiatric diseases were associated with a higher probability of reversion from MCI to NC as assessed with CIRS. This may partially be explained by the complex relationships between aging, depression and cognitive symptoms. The role of depression in the development of dementia in MCI subjects has not been unequivocally established [43, 44]. Depression is common in MCI subjects with a reported prevalence ranging from 35 to 85% [44, 45] and is considered a predictor of progression from MCI to dementia [44]; on the other hand, the impairment in cognitive performance could be due to the depressed mood and not to a neurodegenerative disorder [46]. The subsequent improvement in cognitive performance could be explained by the treatment of the underlying depression disorder. It is noteworthy that major psychiatric disorders were an exclusion criteria from our study and so our results suggest that even mild psychiatric symptoms might have a role in the impairment of cognitive functions.
At a more general level, both physical and mental comorbidities can generate tiredness and lack of concentration, possibly resulting in subjective feeling of cognitive deficits and in eventual impaired cognitive performance [47]. Moreover, several chronic conditions can reduce the amount of the quality of sleep (as previously mentioned), with known reduced cognitive efficiency.
Thus, it is crucial for the clinicians to identify the subjects who suffer from several chronic diseases in order to primarily treat these disorders involved in the cognitive performance of the subjects.
The main strength of the present study is the use of a hard endpoint, as the reversion to NC with a precise neuropsychological assessment and not surrogates of NC, such as only scores in some test of general cognitive performance [10].
The present study has several limitations. First, this is a study reporting preliminary data analysis of the baseline evaluation and was not planned for analyzing the role of comorbidities as time- dependent variable (even though some individuals were followed for more than six years). Moreover, the statistical analysis adopted does not take into account the onset of possible new diseases. We are aware that comorbidities may determine both progression toward dementia or recovering to a normal functional status and without an appropriate time-dependent measure of morbidity, it is difficult to draw any firm conclusion. Nevertheless, it would be anyhow essential to outline some baseline characteristics that could address clinicians to identify those MCI subjects who have a higher probability to revert to a state of NC. Second, the present study had to face with the small sample size related to the APOE genotype information. Nevertheless, it is noteworthy that this study is of current clinical practice and we had to cope with the refusal of most of the subjects to the APOE investigation. At last, the potential selection bias related to the specific clinical setting weakens the possibility of generalizability of results.
For what concerns the future perspective of our study, we will follow up the subjects with annual clinical and neuropsychological evaluation in order to follow the clinical trajectory of the NC condition. It is mandatory to understand whether any diagnosis of MCI might have prognostic implication and if it could still be considered a risk condition for the subsequent development of dementia [48]. Indeed, Roberts and colleagues [49], observed that more than 50% of those MCI subjects who reverted to NC subsequently developed MCI or dementia. Moreover, as a possible future direction of the present study, it could be interesting also to observe the prognostic value of neuropsychological tests both in the reversion to NC and in the progression to dementia.
In line with the findings coming from other memory clinic studies, our results suggest that MCI “construct”, although “unstable” and not unequivocally intended, remains a risk factor for the subsequent development of dementia when used in a memory clinic context. It is crucial to pay special attention to those subjects who suffer from many diseases that can impact to their cognition because their cognitive performance could benefit from preventive strategies aimed at treating the underlying diseases.
