Conversion of Mild Cognitive Impairment to Alzheimer’s Disease in Monolingual and Bilingual Patients

Abstract

Purpose

Conversion rates from mild cognitive impairment (MCI) to Alzheimer’ Disease (AD) were examined considering bilingualism as a measure of cognitive reserve.

Methods

Older adult bilingual (n = 75) and monolingual (n = 83) patients attending a memory clinic who were diagnosed with MCI were evaluated for conversion to AD. Age of MCI and AD diagnoses and time to convert were recorded and compared across language groups.

Patients

Patients were consecutive patients diagnosed with MCI at a hospital memory clinic.

Results

Bilingual patients were diagnosed with MCI at a later age than monolingual patients (77.8 years and 75.5 years, respectively), a difference that was significant in some analyses. However, bilingual patients converted faster from MCI to AD than monolingual patients (1.8 years and 2.8 years, respectively) resulting in no language group difference in age of AD diagnosis. This relationship held after accounting for education, cognitive level, immigration status, and sex.

Discussion

The findings suggest that greater cognitive reserve as measured by language status leads to faster conversion between MCI and AD, all else being equal.

1. Introduction

Mild cognitive impairment (MCI), in which individuals have compromised cognitive ability but no noticeable impairment to everyday functioning, is considered to lie on the trajectory between normal aging and dementia. A meta-analysis of 41 studies that followed MCI patients for more than 3 years found that the annual conversion rate to dementia was approximately 5–10%¹, slightly lower than a widely cited annual conversion rate of 10–15%²; that is, in a random sample of 100 MCI patients, approximately 5–15 will convert to dementia within the first year. Regardless of the true rate, MCI remains a reliable predictor for future conversion to dementia.

Alzheimer’s disease (AD) is the most common form of dementia, comprising approximately 60–70% of dementia cases. In addition to financial stress^3,4, AD has an overwhelmingly negative effect on caregivers and families⁵. It is crucial that future research focuses on understanding the mechanisms by which patients with MCI convert to AD in order to prolong, or possibly halt, conversion and its associated burdens.

Several mechanisms have been proposed to account for ways in which the brain can withstand, or postpone, cognitive decline⁶. Brain reserve (BR), cognitive reserve (CR), and more recently brain maintenance (BM), are three concepts that may help explain preserved cognitive function in the face of aging and neural degeneration. These concepts identify neuroprotective mechanisms or compensatory mechanisms that enable cognitive function to be maintained despite neuropathology. For BR, a quantitative measure of brain structure and volume, such as the number of neurons, synapses, or overall brain size, determines AD incidence. Many studies of BR use generalised brain measures like total intracranial volume, but specific measures such as total pyramidal neurons in the cortex have also been used⁷. This approach has been referred to as a passive threshold model in which cognitive impairment is certain once neuropathology reaches a level that the brain can no longer withstand. However, the term ‘passive’ may be overly simplistic as some studies have demonstrated that active processes such as brain metabolism or experience-related modifications to brain structure also fall under the category of BR^8,9. BM is a complementary concept; whereas BR suggests quantitative brain measures can explain preserved cognitive functioning in the presence of neurodegeneration, BM posits that certain genetic (e.g., allelic variations in genes) or lifestyle factors (e.g., stimulating leisure activities) or their interaction help individuals resist this degeneration in the first place¹⁰.

In contrast to these passive models, CR is an active model; pre-existing cognitive processes actively help the brain cope with pathology. Stern¹¹ posits that AD pathology increases prior to behavioral evidence of impairment, and the degree of CR affects the point at which behavioural deficits appear, dissociating the level of pathology from the appearance of symptoms. Accordingly, those with high levels of CR should be able to fend off negative effects of pathology for longer than those with low levels of CR; that is, individuals with high levels of CR will be coping with greater levels of pathology than those with low CR when behavioral deficits are seen. This leads to the counterintuitive proposal that high CR individuals convert from MCI to AD faster than those with low CR (Figure 1; see also Valenzuela¹²). Because environmental and lifestyle factors influence levels of CR, two individuals with similar brain volumes and levels of brain pathology may exhibit dissimilar levels of impairment based on differences in CR. Among the factors posited to increase CR are education and higher occupational achievements¹¹, as well as musical expertise^13,14,15. Other activities such as video-game playing¹⁶ and taxi-driving¹⁷, while not explicitly linked to CR, have also been proposed as means to improve cognitive functioning.

Figure 1.

Change in memory function over time for individuals with low and high cognitive reserve. Taken from Stern¹¹.

Another lifestyle factor, bilingualism, has also been posited to increase CR. Of all activities with neuroplastic benefits, language use is the most sustained, consuming the largest proportion of time within a day. It also activates regions across the entire brain, including frontal, temporal, parietal, and some posterior regions¹⁸. As such, it has been suggested that the experience of bilingualism has effects on brain regions and processes beyond language processing to include nonverbal cognitive performance¹⁹. Bilingualism is therefore a strong candidate to increase CR and thus delay the age of onset of AD symptoms in elderly patients.

Bialystok et al.²⁰ investigated the effect of bilingualism on age of onset of dementia. In their sample of 184 patients, bilinguals showed symptoms of dementia on average 4.1 years later than monolinguals, and this delay was greater (4.3 years) when considering only onset of AD and not other dementias. This finding has been replicated^21,22, but there were concerns that the protective effect of bilingualism could be explained by immigration status rather than bilingualism²³ (see Schweizer et al.²⁴ for a rebuttal). Importantly, therefore, Alladi et al.²⁵ replicated the original findings in a large patient population in India in which bilingualism was not tied to immigration. Bilinguals in this study were on average 4.5 years older than monolinguals when diagnosed with dementia. Using a different approach, Klein et al.²⁶ reported that AD incidence rates in 93 countries significantly declined as population multilingualism increased, a relation that became stronger when life expectancy was included in the model.

Thus, bilingualism appears to be a source of CR that postpones appearance of AD symptoms, but to our knowledge no studies have investigated the effect of bilingualism on time to convert from MCI to AD. Understanding the time course of conversion has substantial implications for those with MCI, particularly with respect to a patient’s – and their family’s – quality of life. If bilingualism increases levels of CR in those with neuropathology, then according to Stern’s model¹¹, bilingual patients with MCI should convert faster to AD than their monolingual peers, all else being equal. Taken together, one would expect later onset of MCI but faster conversion from MCI in bilingual patients than in monolingual patients.

The current study addresses this hypothesis. Participants were older adults who had been referred to a memory clinic because of cognitive complaints and received a diagnosis of MCI. It was assumed that bilingualism increases levels of CR, and therefore bilingual patients would be better able to withstand the damaging effects of neuropathology than their monolingual peers. Therefore, MCI should be diagnosed at an older age for bilingual patients, but because higher CR is associated with increased neuropathology, they would also experience a more rapid rate of decline than found for those with low CR¹¹. Thus, the prediction was that bilingual patients will be older than monolingual patients at the time of MCI diagnosis yet will convert to AD faster than comparable monolingual patients. Other factors that have been shown to play a role in the development or diagnosis of AD are also considered, namely, formal education^27,28, immigration²⁹, sex²⁰, and genetics³⁰.

2. Method

We examined consecutive records of all patients referred to the Sam and Ida Ross Memory Clinic at Baycrest Health Sciences in Toronto, Canada, for cognitive complaints between 2007 and 2017. Patients undergo medical history, physical and mental status examinations including the Mini-Mental State Examination (MMSE) or Montreal Cognitive Assessment (MoCA), in addition to CT, SPECT, MRI, and blood screening, when possible. Patients additionally underwent a neuropsychological evaluation that included clinical interview and the administration of a 3-hour test battery consisting of subtests from the Kaplan-Baycrest Neurocognitive Assessment (Word Lists, Complex Figure, Clocks, Spatial Location Memory, Practical Problem Solving), Wechsler Memory Scale-Revised (Logical Memory), Wechsler Abbreviated Scale of Intelligence (Vocabulary, Similarities, Matrix Reasoning), Wechsler Adult Intelligence Scale −3 (Digit Span), Delis-Kaplan Executive Function System (Colour-Word Interference), Trail Making Test, Verbal Fluency, Boston Naming test, and mood questionnaires. Patients’ records were examined for a diagnosis of MCI made by agreement between the attending physician and neuropsychologist, either at initial appointment or follow-up, with subsequent conversion to AD while attending the Memory Clinic. Patients whose cognitive deficits may have been a result of or influenced by a concurrent ailment, such as traumatic brain injury, cancer, or chronic depression, were excluded. The final sample consisted of 158 patients, 119 of whom had a diagnosis of MCI with memory impairment (amnestic MCI, or aMCI) prior to conversion to AD, 33 with a diagnosis of MCI without explicit acknowledgement of memory difficulty, and 6 with a diagnosis of MCI with a vascular component.

During the intake interview and subsequent neuropsychology appointment, patients and significant others provided information related to language status such as languages spoken, self- and family-reported fluency, place of birth, and year of immigration. Knowledge of more than one language, regular use of a language in addition to English, and living in a non-English speaking country were the primary determinants to establish bilingualism. Any patients who failed to provide sufficient information to determine language status were not included in the study. Thus, 83 patients were classified as monolingual and 75 were bilingual. Immigration was a factor for both monolinguals (n = 17; 20.5%) and bilinguals (n = 47; 62.7%), and occurred in the 1950s (n = 13), 1960s (n = 16), and 1970s (n = 10). The bilinguals included speakers of 23 non-English languages including Yiddish (n = 21), French (n = 10), Hebrew (n = 6), German (n = 5), and Romanian (n = 5), among others (n = 28).

Age of onset of cognitive complaints was established during the intake interview with the patient and family in response to such questions as when they first noticed a decline in cognitive function and how this decline presented itself. Previous research showed that monolingual and bilingual patients did not differ regarding how long they waited before seeking help for their complaints²⁰. Age of onset of MCI was set as the age the patient was first diagnosed with MCI by the attending physician. Diagnosis was made using the NIA-AA core clinical criteria for MCI³¹. A similar method was used to establish age of onset of AD when the physician diagnosed a conversion to AD based upon core clinical criteria for probable AD³². Development of MCI and AD occurred between appointment dates with the clinicians, but both monolingual and bilingual patients returned to the clinic every 3–12 months for follow-up with an average of approximately 2 appointments per year.

Magnetic resonance imaging (MRI) scans were available for 24 monolinguals and 24 bilinguals to determine if there were neural correlates to the variables of interest. However, these scans were conducted for diagnostic purposes from a variety of hospitals using different MRI machines. As a result, the wide variability between hospital imaging procedures could not be controlled, and the low scan quality was unsuitable for performing extensive analyses.

3. Results

The mean values for the variables of interest are presented in Table 1. There was no significant group difference in age of diagnosis of MCI, F(1, 156) = 3.03, p = .08, η² = .02, or AD, F(1, 156) = 1.23, p = .27, η² = .01, but bilinguals converted from MCI to AD faster than monolinguals, F(1, 156) = 8.81, p = .003, η² = .05. A density plot of this effect is shown in Figure 2.

Table 1.

Mean values (and standard deviations) for all measures by language group further divided by sex and immigration status where available.

Language group	Subset	N	Education (years)	MMSE at MCI (/30)	Symptom onset prior to first visit (years)	Age of MCI diagnosis (years)	Age of AD diagnosis (years)	Years from MCI dx to convert
Monolingual	Total (by sex)	83	14.9 (3.5)	26.9 (2.4)	2.5 (2.0)	75.5 (7.6)	78.2 (7.7)	2.6 (1.8)
	Male	41	15.9 (3.9)	27.5 (1.7)	2.7 (2.0)	76.8 (6.7)	79.7 (7.0)	2.9 (1.9)
	Female	42	14.0 (2.8)	26.3 (2.8)	2.4 (2.0)	74.2 (8.1)	76.6 (8.0)	2.4 (1.6)
	Total (by immigration)	68	14.9 (3.6)	26.8 (2.5)	2.7 (2.1)	75.4 (8.0)	77.9 (8.2)	2.5 (1.7)
	Immigrant	17	14.2 (4.0)	25.9 (3.4)	2.1 (1.7)	75.7 (8.7)	78.8 (8.6)	3.0 (2.1)
	Non-Immigrant	51	15.1 (3.5)	27.1 (2.1)	2.9 (2.2)	75.3 (7.9)	77.6 (8.1)	2.3 (1.5)
Bilingual	Total (by sex)	75	13.9 (3.6)	25.9 (3.0)	2.7 (1.8)	77.8 (8.2)	79.8 (8.4)	1.9 (1.7)
	Male	36	14.9 (3.7)	26.1 (2.7)	2.9 (2.0)	78.9 (7.3)	80.7 (7.6)	1.8 (1.4)
	Female	39	12.9 (3.3)	25.6 (3.3)	2.4 (1.5)	76.6 (9.0)	78.5 (8.8)	1.9 (1.6)
	Total (by immigration)	72	13.8 (3.7)	25.8 (3.0)	2.7 (1.8)	77.8 (8.3)	79.6 (8.4)	1.8 (1.5)
	Immigrant	47	13.3 (3.5)	25.1 (3.2)	2.5 (1.6)	76.4 (7.5)	78.2 (7.5)	1.8 (1.6)
	Non-Immigrant	25	14.7 (3.9)	27.4 (1.5)	3.0 (2.2)	80.4 (9.3)	82.4 (9.3)	2.0 (1.4)

Figure 2.

Density plot of time to convert from MCI to AD in monolingual and bilingual patients.

Other factors were then considered for their role in the age of diagnosis and conversion time, although in some cases, data were available for only a subset of the sample. From the 158 patients, 125 completed an MMSE at the appointment in which they were diagnosed with MCI. Monolinguals had higher MMSE scores than bilinguals, F(1, 123) = 4.27, p = .04, η² = .03, but because of issues with the MMSE as a diagnostic tool, MoCA scores were also examined for the smaller sample of patients for whom these scores were available (further explained in the Discussion). Thirty-six monolinguals and 38 bilinguals had completed a MoCA, with an average score of 22.4 for monolinguals and 21.3 for bilinguals, a difference that was not statistically significant, F(1, 72) = 1.75, p = .19, η² = .02.

Formal education also contributes to cognitive reserve and could therefore delay the development of AD. However, the groups did not differ in years of education, F(1, 155) = 3.32, p = .07, η² = .02, so no further analyses with education were conducted.

To determine if immigration was a factor, three 2-way ANOVAs for language group and immigration were run for each of age of MCI diagnosis, AD diagnosis, and conversion time. Patients for whom immigration status was undetermined (e.g., country of birth was unknown) were excluded from the analyses, producing a sample of 68 monolinguals (17 immigrants) and 72 bilinguals (47 immigrants). Descriptive statistics for this subset are included in Table 1. For age at MCI diagnosis, there was no effect of immigration status, F(1, 136) = 1.78, p = .18, η² = .01, but monolinguals were diagnosed at an earlier age than bilinguals, F(1, 136) = 4.48, p = .036, η² = .03. For age at AD diagnosis, neither immigration status, F(1, 136) = 1.47, p = .23, η² = .01, nor language group, F(1, 136) = 2.63, p = .11, η² = .02, were significant. Finally, the analysis of time to convert found no effect of immigration status, F < 1, but a significant effect of language group, F(1, 136) = 6.10, p = .01, η² = .04, with shorter conversion times for bilinguals. There were no significant interactions between immigration and language group in any of the analyses, Fs < 3.1.

Given that men may postpone clinic visits longer than women when symptoms of memory loss appear¹⁹, analyses examining time between symptom onset and first clinic visit were run using sex and language group as factors. The time delay showed no effect of sex, F(1, 153) = 1.69, p = .20, η² = .01, language group, F < 1, or their interaction, F < 1. However, for age of MCI diagnosis, men were diagnosed at an older age than women, F(1, 154) = 3.91, p = .05, η² = .02, with no effect of language group, F(1, 154) = 3.16, p = .08, η² = .02, or interaction of sex and group, F < 1. Similarly, men were older than women when diagnosed with AD, F(1, 154) = 4.47, p = .04, η² = .03, but there was no effect of language group, F(1, 154) = 1.32, p = .25, η² = .01, or interaction, F < 1. However, for conversion times, there was only an effect of language group in which bilinguals converted faster than monolinguals, F(1, 154) = 8.73, p = .004, η² = .05, and no effect of sex, F < 1, or interaction effect, F < 1.1.

The final factor was the role of an AD genetic component. Two-way ANOVAs were used to examine familial history of AD and language group on age of diagnoses and time to convert. One patient was missing information about family history, so 82 monolinguals (32 with positive history) and 75 bilinguals (29 with positive history) were included. For MCI diagnosis, there was no effect of family history, F < 1, language group, F(1, 153) = 3.22, p = .07, η² = .02, or their interaction, F(1, 153) = 2.03, p = .16, η² = .01. Similarly, for age of AD diagnosis there was no effect of family history, language group, or their interaction, Fs < 1.4. In contrast, an ANOVA examining the diagnosed time to convert revealed a significant interaction between language group and family history, F(1, 153) = 4.64, p = .03, η² = .03. A post hoc Tukey test showed that bilinguals without a family history of AD converted significantly faster (M = 1.7 years) than monolinguals without a family history of AD (M = 3.0 years), p = .002, but there was no difference between bilinguals and monolinguals with family histories of AD on conversion time, t < 1. No other pairwise comparisons were significant, ps > .05.

Because there were significant differences between language groups for both age of MCI diagnosis and conversion time, it was possible that older age of diagnosis was associated with a faster time to convert. However, a correlation between age of MCI diagnosis and time to convert was not significant, r(156) = −0.09, p = .25. Similar results were found when examining this correlation separately by language group, for both monolinguals, r(82) = −0.06, p = .59, and bilinguals, r(74) = −0.07, p = .56.

4. Discussion

Bilingual patients converted faster from MCI to AD than monolinguals, a difference that was significant (1.9 years and 2.6 years, respectively). In spite of this, monolingual and bilingual patients had equivalent levels of education and initial scores on the MoCA, although MMSE scores were slightly higher for monolinguals. This pattern is consistent with Stern’s³³ claim that high CR leads to faster conversion times once a clinical threshold has been passed, supporting our description of bilingualism as a source of CR.

Contrary to previous research^20,25 and our predictions, bilingual and monolingual patients did not differ in age at time of AD diagnoses. Regarding age of diagnosis of MCI, bilinguals were older than monolinguals but the difference in age was only significant for the subset of patients for whom immigration status was available. This result is in line with previous research³³ that showed a later age at diagnosis of single-domain aMCI for bilinguals than monolinguals. As our study did not explore the distinction between MCI subtypes, it is hard to make direct comparisons to this previous finding. However, the conversion time from MCI to AD was always significantly shorter for bilinguals, so it is not surprising that the age of AD diagnosis would converge for the two language groups and thus show no difference.

One concern is that previous studies have found a difference between language groups on age of diagnosis of AD^20,35,36 (see Chertkow et al.²¹ for a small but not significant effect), a difference that was not found in our study. However, studies have shown a range of ages at diagnosis. Across studies, monolinguals have been diagnosed with AD from an average age of 74.2 years³⁵ to 78.2 years, whereas the range found for bilinguals is 77.6 years¹⁹ to 81.4 years³⁵. As such, the midrange value is around 76.2 years for monolinguals and 79.5 years for bilinguals across studies; in the present study these figures were 75.5 years for monolinguals and 77.8 years for bilinguals, easily within this range. In this sense, the absence of a statistical difference between ages of AD diagnosis between groups in the current study is not necessarily contrary to previous research, but rather reflects the range and variance in these measures and contributes to the overall picture. Future studies with a larger sample size and more precision regarding stages and subtypes of MCI prior to conversion to AD may help tease apart this discrepancy in findings.

To our knowledge, this is the first study to investigate conversion times from MCI to AD in monolingual and bilingual patients. Previous research has examined the onset of MCI³⁴ or Alzheimer’s symptoms between these two groups, but not the time course of progression from MCI to AD. The findings are important for several reasons. First, the time course of conversion is indicative of disease progression. Different conversion rates between language groups intuitively indicate different rates of progression, and the present results show a faster rate of decline for bilinguals than monolinguals. As suggested by Stern¹¹ this would imply that bilinguals have higher levels of CR than monolinguals, but also greater levels of neuropathology when the decline begins and presumably when the diagnosis is made. Compared to monolinguals, bilinguals’ higher level of CR was arguably associated with greater levels of neuropathology at time of MCI diagnosis, faster cognitive decline, and subsequent conversion to AD.

Additionally, this difference in conversion times has implications for the diagnosis and prognosis of MCI. Several risk factors have been identified for MCI including (but not limited to) age, sex, genetic factors, and education¹. Given the findings of the present study, we posit that language use is another such factor. The current findings suggest that bilinguals could be expected to be diagnosed with MCI later in life than their monolingual peers when also accounting for sex and immigration. Once diagnosed, however, the prognosis for disease progression is notably poorer.

Because MMSE scores were statistically different at the time of MCI diagnosis, it might be argued that bilinguals were already more cognitively impaired than monolinguals, explaining their faster conversion to AD. However, this difference of one point (26.9 and 25.9 for monolinguals and bilinguals, respectively) is not clinically relevant despite being statistically significant. MMSE has a ceiling effect and the limited performance range for cognitively healthy individuals increases the probability that predementia individuals will fall into the normal range of 24 or above³⁷. Both groups are within the normal range and would be classified as such for purposes of diagnosis. Due to its flaws, some have argued that the MMSE is limited in its capability to detect MCI³⁸, or that it is less sensitive than other screening measures such as the MoCA³⁹. With this in mind, scores on the MoCA may be a better indicator of cognitive health. Compared to the MMSE, subjects’ scores on the MoCA fall below the healthy cut-off of 26, and even the recently suggested score of 23⁴⁰, indicating a diagnosis of MCI. Given the criticisms of MMSE and higher sensitivity of the MoCA, we suggest that group differences on the MMSE are not indicative of MCI age or conversion time to AD.

This study provides support for the proposal that bilingualism is a factor that contributes to CR. The difference in conversion times from MCI to AD suggests that neural degradation may have been more advanced in bilinguals, leading to faster conversion once a threshold was passed. Therefore, clinical assessments should consider the role of language status to guide diagnoses and treatment.