Introduction
In clinical practice, practitioners are increasingly confronted with bilingual patients referred for cognitive complaints. This frequently raises the issue of the effect of the language used to assess cognitive abilities on the determination of cognitive status and ultimately the diagnosis.
The current literature in neuroscience and linguistics offers various definitions of bilingualism; however, the one that seems most fitting for our study describes it as the ability to switch between two languages, suggesting that a bilingual individual “utilizes at least two languages in their daily life.” Reference Grosjean1 Numerous factors, both intrinsic (age of acquisition, proficiency level, learning context, etc.) and extrinsic (linguistic practice, usage contexts, social hierarchy, etc.), influence bilingual competencies and allow for the determination of the type of bilingualism. Reference Geiger-Jaillet2,Reference Rezzoug, Sprenger-Charolles and Colé3 For instance, the age of first exposure distinguishes between early and late bilingualism: early bilingualism can be simultaneous or sequential – simultaneous if both languages are acquired from birth and sequential if both languages are acquired from an early age but successively. Reference De Houwer and A.4,Reference Hamers and Blanc5 Moreover, bilingualism is considered late when the second language is learned after the age of 12 years. Reference De Houwer and A.4,Reference Heredia and Cieślicka6 Distinctions are also made between prominent bilingualism, where the individual has a better mastery of one language over the other, and balanced bilingualism, where the individual has similar competencies in both languages. Reference Birdsong7–Reference Kohnert9
Accordingly, bilingualism influences neuropsychological test performance in older adults, and this has consequences on the assessment of language and executive functions. Reference Bialystok, Poarch, Luo and Craik10,Reference Celik, Kokje, Meyer, Frölich and Teichmann11
As a matter of fact, bilingualism may result in slower lexical access and lower scores in naming and verbal fluency tasks Reference Celik, Kokje, Meyer, Frölich and Teichmann11–Reference Rosselli, Ardila and Araujo16 due to language interference. Reference Sadat, Martin, Alario and Costa17–Reference Zeng, Kalashnikova and Antoniou19 However, associated with favorable living conditions, bilingualism is believed to have positive cognitive effects, Reference Bialystok20–Reference Treccani and Mulatti22 particularly on measures assessing inhibitory control, Reference Celik, Kokje, Meyer, Frölich and Teichmann11 whether through conflict resolution tasks evaluated by the Stroop test Reference Bialystok, Poarch, Luo and Craik10,Reference Kousaie23,Reference Massa, Köpke and El Yagoubi24 or flexibility and switching in adults. Reference Abutalebi and Green25 Nevertheless, other studies have failed to demonstrate the tangible benefits of bilingualism in the area of executive functions. Reference Anderson, Saleemi and Bialystok12,Reference Paap and Greenberg26,Reference von Bastian, Souza and Gade27 Consequently, norms for monolinguals on executives and language tests should be adjusted when assessing bilingual individuals in their non-prominent language to avoid misestimating their abilities and to ensure a reliable diagnosis.
In patients with dementia, especially in Alzheimer’s disease (AD), it is frequently claimed that language impairment is more marked in the non-prominent language Reference Ivanova, Salmon and Gollan28–Reference Mendez, Perryman, Pontón and Cummings31 although this remains controversial. Reference Gollan, Salmon, Montoya and da Pena32–Reference Salvatierra, Rosselli, Acevedo and Duara34 In assessing naming abilities, a recent study highlighted the heightened sensitivity of using the prominent language in bilingual AD patients. Reference Gollan, Stasenko and Salmon35 This was evidenced through the application of the Multilingual Naming Test. Reference Gollan, Stasenko and Salmon35
Regarding screening tests, only two studies were identified that investigate the influence of language on cognitive assessment outcomes in bilingual individuals, highlighting the complexities of language choice in the Montreal Cognitive Assessment (MoCA) and the Mini-Mental State Examination (MMSE). One study Reference Briceño, Mehdipanah and Gonzales36 explored the MoCA Reference Nasreddine, Phillips and Bédirian37 and revealed that the choice of language used in the MoCA significantly influenced the cognitive test outcomes among balanced bilinguals. As a matter, in this study, balanced bilinguals that completed the MoCA in English performed better than balanced bilinguals that completed the MoCA in Spanish. However, its design does not compare scores in the two languages in unbalanced bilinguals and exploration was only done in healthy controls (HC). As for the MMSE, Reference Folstein, Folstein and McHugh38 only one study was found and compared both language scores – Irish and English – in an Irish prominent population (based on informal measure) and showed better sensitivity in the prominent and societal majority language for the diagnosis of cognitive impairment. Reference Ní Chaoimh, De Bhaldraithe and O’Malley39 However, authors used one single cognitive test, and bilingualism subtype, based on an informal report, was not clearly specified. Moreover, some may question the similarity between the languages’ tested in this study, being Irish (i.e., Goidelic language) and English (i.e., Germanic language). The phenomenon of crosslinguistic influence highlights how similarities between languages can aid in testing in another language, potentially facilitating cognitive assessments.
Overall, controversies between studies in bilingual populations globally and in AD more precisely are likely to be due to the use of variable determination of bilingualism subtype and the lack of control for the bilingualism subtype on performance across languages.
Moreover, it is often challenging to conduct assessments of bilingual individuals in their first and most proficient language. As a result, testing is commonly carried out in the societally majority language, which may not be the individual’s primary language. Therefore, there are concerns about tests validity and the influence of the used language on cognitive screening accuracy. It’s pertinent to note, especially in European contexts influenced by non-European immigration, that patients are frequently tested in a European language, usually their second language (i.e., non-prominent).
Lebanon offers a unique perspective for studying bilingualism due to its multilingual environment. The country utilizes several languages, including Modern Standard Arabic (the official language used mainly in writing and media), Lebanese Arabic (spoken by 93.7% of the population for daily communication) Reference Leclerc40 French, English and Armenian. Reference Hoteit41–Reference Saliba44 A survey in 2007 revealed that 45% of the Lebanese population spoke French and 40% English, Reference Gingras45 leading to the frequent use of Lebanese, French and English within the same sentence. Reference Kotob46
This prevalent bilingualism is also reflected in the education system. In the Lebanese national curriculum, subjects such as Arabic, geography, history and civic education are taught in Arabic, while other subjects like physics, chemistry, biology, science and foreign languages are taught in either English or French. Reference Bayloun47 This system underscores the simultaneous or sequential bilingualism present in the Lebanese society, offering insights into balanced versus dominant bilingualism in a context where mastering at least two languages is commonplace despite the multilingual setting. In the context of Lebanon’s multilingual environment, the MMSE is particularly utilized for cognitive assessment due to its calibration for the Lebanese population.
This brief paper investigated the effect of the language used for administering MMSE on its ability to accurately diagnose cognitive impairment among Lebanese bilinguals (Arabic-French). This study leveraged a previously validated approach for categorizing bilingual subtypes, utilizing the Language Experience and Proficiency Questionnaire (LEAP-Q) Reference Marian, Blumenfeld and Kaushanskaya48 and additional language tests. Reference Kassir, Abboud and Godefroy49
Population and methods
The main inclusion criteria of patients were the following: bilingual (Arabic and French) patients aged between 55 and 92 years referred between January 10, 2020 and November 11, 2021 to neurology department of Hôtel-Dieu de France in Beirut, Lebanon, with mild to moderate dementia (defined by ≥ 16 in both languages (Arabic and French) fitting the criteria for probable AD Reference McKhann, Drachman and Folstein50 consenting to participate to the study and free of exclusion criteria (illiteracy, any previous psychiatric or neurological disease affecting cognition and any perceptual [auditory and/or visual] or motor deficit precluding cognitive testing) with no severe comprehension difficulties (Token test Reference De Renzi and Faglioni51 in each language ≥ 26). MMSE was administered in Lebanese Arabic Reference El-Hayeck, Baddoura and Wehbé52 and French Reference Kalafat, Hugonot-Diener and Poitrenaud53 in counterbalanced order. Both MMSE assessments were conducted within the same assessment session following the administration guidelines specified by the respective authors.
These criteria encompassed 29 AD patients (20 females, mean age = 77.2 ± 5.9), 26 with mild (as determined by the worst MMSE score ≥ 20) and 3 moderate dementia (worst MMSE ≥ 16) (Table 1). The HC group consisted of healthy Lebanese resident participants who fulfilled the following inclusion criteria: (1) must have been exposed before the age of 12 years to at least 2 languages, Arabic and French, (2) aged between 55 and 92 years, (3) living in Lebanon and (4) free of exclusion criteria (illiteracy, any declared psychiatric or neurological disease affecting cognition and any auditory or visual or motor deficit precluding cognitive testing).
Table 1. Demographic characteristics of healthy controls group (HC) (n = 93) and Alzheimer’s disease (AD) patients’ group (n = 29) expressed as number (n) of participants and mean ± standard errors
Ar-Fr dMMSE: MMSE score in Arabic – the score in French.
Level (1): less or equal to 8 years of education after the end of elementary school; (2): between 8 and 11 years of education after the end of elementary school; (3): baccalaureate level or equivalent and above.
Bilingualism subtype was determined using the previously validated bilingualism index Reference Kassir, Abboud and Godefroy49 based on proficiency level in speaking and understanding of the LEAP-Q Reference Kassir, Abboud and Godefroy49 informed by the patient and caregiver. This subdivided the AD group into 8 Arabic-prominent and 21 balanced patients. Their performance was compared to those of a control group of 93 bilingual HC Reference Kassir, Roussel, Abboud and Godefroy54 (54 females; mean age = 67.99 ± 9.3) with 35 Arabic-prominent and 58 balanced HC. The agreement of the ethics committee of the Hôtel-Dieu de France Hospital has been granted for the study (file CEHDF 1449).
Statistics
Statistical analyses were based on a validated framework for the analysis and interpretation of cognitive data. Reference Godefroy, Gibbons and Diouf55 First, MMSE scores were adjusted for age and education separately in each language. Scores were adjusted for age and education level using a regression analysis and coefficients computed in HC. Residuals were used in all analyses.
MMSE scores across groups and languages
Adjusted MMSE scores (adjMMSE) were compared across groups and languages using a repeated analysis of variance (ANOVA) with the between-subject factors group (HC, AD) and bilingualism subtypes (Arabic prominent, balanced) and the within-subject factor language (Arabic, French). Then, we examined whether the differential decrement across languages was related to severity of cognitive impairment by exploring the correlation between the best MMSE score (i.e., disease severity) and the difference of MMSE scores across languages (Ar-Fr dMMSE) (i.e., the score in Arabic – the score in French).
Discriminative ability of adjusted MMSE scores
Ability of adjMMSE to discriminate AD from HC was examined in each language (Arabic, French) by a stepwise logistic regression analysis with the group (HC, AD) as dependent variable. The independent variables submitted to analysis were bilingualism subtype (Arabic prominent, balanced) and adjMMSE scores in each language (Arabic, French). This analysis was repeated in each subgroup (Arabic prominent, balanced).
All statistical analyses were performed using SPSS®. A p value ≤ 0.05 was considered significant, unless otherwise indicated.
Results
MMSE scores across groups and languages
MMSE scores (Table 1) differed across groups (F (1,118) = 63.8; p <.001) due to overall lower scores in AD group (HC group: 27.3 ± .26; AD group: 22.9 ± .5). MMSE did not differ according to bilingualism subtypes (F (1,118) = 1.1; p = .29) (Arabic prominent: 24.8 ± .5; balanced: 25.3 ± .3) and to language (F (1,118) = .002; p = .96) (Arabic: 25.45 ± .3; French: 24.7 ± .3). Interactions did not reach significance (language × bilingualism subtype: p = .14; language × group: p = .3; bilingualism subtype × group: p = .8; language × bilingualism subtype × group: p = .35).
This analysis indicates that MMSE scores were lower in AD patients regardless of the language used and the bilingualism subtype.
The best MMSE score across languages did not correlate (R 2 = −0.07, p = .4) with the Ar-Fr dMMSE even after controlling for groups (R 2 = −0.04, p = .6). Thus, the differential decrement across languages was not related to the severity of cognitive impairment.
Discriminative ability of adjusted MMSE scores
The stepwise logistic regression selected the Arabic adjMMSE score (odds ratio [OR]:0.54, 95%confidence interval [CI]: 0.43–0.68, p = .0001) to discriminate AD from HC and the bilingualism subtype was not significant (p = .3). Accordingly, similar results were obtained in repeated analyses performed in both the Arabic prominent (Arabic adjMMSE score) (OR:0.44, 95%CI: 0.26–0.75, p = .003) and the balanced subgroups (OR:0.58, 95%CI: 0.45–0.74, p = .0001).
Discussion
This study shows that MMSE is impaired in bilingual AD patients to a similar extent whatever the language used. More importantly, it shows that cognitive screening using MMSE in Arabic provides the most discriminative measure of cognitive impairment in both Arabic-prominent and balanced Lebanese bilinguals. Regarding Arabic-prominent patients, this indicates that the prominent language (i.e., Arabic) provides the most sensitive measure of impairment. Regarding balanced Lebanese bilinguals, our results indicate that the use of the societal majority language (i.e., Arabic) provides the most sensitive measure of impairment.
This raises questions about cognitive screening methodologies for immigrants/foreigners, potentially affecting their assessments and diagnosis.
Cognitive impairment across languages
Overall, our results contrast with previous studies suggesting that cognitive impairment on MMSE of non-balanced bilinguals is more severe in the non-prominent language. Reference Ivanova, Salmon and Gollan29–Reference Mendez, Perryman, Pontón and Cummings31 The discrepancy with previous results might be due to the assessment of various bilingual subtypes defined according to a strict and validated methods. Reference Kassir, Abboud and Godefroy49 It might also be due to the analysis of MMSE scores after adjustment for age and education in each language. Reference Godefroy, Gibbons and Diouf55 This adjustment for demographic factor is critical as balanced and unbalanced bilinguals are likely to differ regarding age and education, a characteristic that was not observed in our population.
Sensitivity of language use in cognitive screening
Overall, our results are congruent with Ní Chaoimh Reference Ní Chaoimh, De Bhaldraithe and O’Malley39 et al. (2015) and indicate that screening for cognitive impairment using the MMSE is more sensitive when the prominent language is used in unbalanced bilinguals or when the societal majority language is used in balanced bilinguals. This might be attributed to greater daily life exposure and its practical use, although this warrants further studies.
In light of these considerations, we have elaborated in our article that the observed sensitivity of the MMSE in Arabic-prominent individuals, which was more pronounced when tested in their prominent language, suggests a complex interplay between language proficiency and cognitive assessment outcomes.
This finding holds clinical significance, suggesting that employing the standard method for detecting cognitive impairment in foreigners/immigrants (i.e., use of the non-prominent language), such as utilizing the non-prominent language, unveils a subtle reduction in the sensitive measure on the MMSE. In practical terms, we recommend employing the prominent language for cognitive screening in dominant bilinguals and the societal majority language for balanced bilinguals. These recommendations aim to bridge the gap between research findings and clinical practice, facilitating more nuanced and culturally sensitive approaches to cognitive screening in bilingual individuals. The rich linguistic and cultural contexts of bilingual individuals must be taken into consideration to enhance the reliability and validity of cognitive assessments.
Study limitations and future directions
Several limitations of this study should be acknowledged. First, the sample size of the patient group is relatively modest, and this may lead to underestimate the crossed effects of language and bilingualism subtype in the ANOVA. Although this limitation holds, the sample size did not prevent demonstrating a language advantage (i.e., Arabic) to discriminate AD patients from HC. Second, our cross-sectional design does not rule out a different temporal course of impairment in the two languages with the progression of AD. As a matter of fact, in a longitudinal exploration, Reference Ivanova, Salmon and Gollan29 authors highlighted a larger deficit for the prominent language at the initial stage of the disease together with a steeper deterioration in the non-prominent language on the longitudinal assessment on the Boston Naming Test Reference Kaplan, Goodglass and Weintraub56 in 12 unbalanced Spanish-English bilinguals with probable AD. An exploration using the MMSE remains unexplored across languages. Although a longitudinal study is mandatory to document such difference across languages in the time course of cognitive performance, the lack of correlation between the severity of cognitive impairment and the differential MMSE impairment across languages provides no evidence for such interpretation in mild to moderate AD population. Third, due to the pandemic and Lebanese situation, the present study was not able to include a subgroup of French-prominent AD bilinguals of a size necessary to be informative. Thus, we were unable to determine whether French MMSE would be more sensitive (than Arabic MMSE) to detect cognitive impairment in French-prominent AD. Although this limitation does not prevent from showing a clear effect in Arabic-prominent and balanced bilinguals, the next step will be to include these patients.
More broadly our goal of future studies is to assess bilingual patients with the three bilingualism subtypes from different countries and languages in order to assess the cross-cultural generalizability of the present findings. Notably, we envision the inclusion of a group of bilingual Alzheimer’s subjects, particularly those proficient in French: collaborative efforts with French centers will be sought to facilitate the collection of pertinent data, contributing to the comprehensive nature of our study and further enhancing its relevance. This will be necessary to optimize the diagnosis of cognitive impairment in bilinguals which constitutes a growing concern. Additionally, it would be beneficial to explore in future studies whether the advantage observed with the MMSE in our current research extends to other pathologies. Such an investigation could further contribute to our understanding of cognitive assessments in diverse linguistic and clinical contexts, potentially leading to more nuanced and effective diagnostic tools.
Data availability
The data that support the findings are stored at Laboratoire de Neurosciences Fonctionnelles et Pathologiques (LNFP) UR UPJV 4559 in Amiens (France) and may be shared by the corresponding author upon reasonable request.
Acknowledgments
We wish to thank all the participants who volunteered in this study for their time and effort in completing our tests.
Author contributions
R.K. conceived and designed the study, analyzed the data and wrote the manuscript. M.R. contributed to the study design, provided critical feedback on experiments and revised the manuscript.
H.A. contributed to the study design and revised the manuscript.
O.G. conceived and designed the study, analyzed the data, provided critical feedback and revised the manuscript.
Funding statement
The study was supported by the Hubert Curien CEDRE program implemented in France by the Ministry of Europe and Foreign Affairs (MEAE) and the Ministry of Higher Education, Research and Innovation (MESRI) and in Lebanon by the Ministry of Education and Higher Education and by L’Oréal-UNESCO For Women in Science Levant Young Talents Program.
Competing interests
Rania Kassir, Martine Roussel, Halim Abboud and Olivier Godefroy declare no conflict of interest.
