A short and simple bedside test to detect cognitive fluctuations in patients with dementia with Lewy bodies

The establishment of cognitive fluctuations is important when dementia with Lewy bodies (DLB) is suspected but can be especially difficult in the absence of a caregiver who lives with the patient. We examined the possibility of using fluctuating scores on a forward (FDS) and a backward digit span (BDS) test as a marker for cognitive fluctuation. Patients with DLB (21), other forms of dementia (14 with Alzheimer’s disease, 8 with vascular dementia) and 20 controls were asked to perform an FDS and BDS twice, with an interval of 20 min. Seventy percent of patients with DLB showed evidence of cognitive fluctuations for at least one test, while less than 10% of controls and patients with other dementias did. Evidence of cognitive fluctuations on at least one of both tests classified 83% of patients correctly (i.e. DLB or not), with a sensitivity of 70% and a specificity of 90%. Repeated forward and backward digit span tests seem a valid, short, easy and inexpensive bedside tool to detect cognitive fluctuations in the diagnostic work-up of DLB, even in the absence of a caregiver, which limits the use of questionnaires.


Introduction
Cognitive fluctuations are spontaneous alterations of cognitive abilities over a short time, often accompanied by disturbances in alertness or arousal [1]. According to the fourth revised diagnostic consensus criteria, their presence is one of the 4 core clinical criteria for the diagnosis of probable dementia with Lewy bodies (DLB), together with visual hallucinations, parkinsonism and Rapid Eye Movement (REM)sleep behavior disorder [2]. It is DLB's most frequent and specific symptom, touching 80-90% of all patients and only 20-25% of patients with Alzheimer's disease (AD) [3].
Probable DLB is diagnosed when a patient has dementia with at least two of these 4 core clinical criteria or only one core clinical criterion together with at least one of the following indicative biomarkers: reduced dopamine transport in basal ganglia as seen on Single Photon Emission Computerize Tomography or Positron Emitting Tomography, low uptake on 123-I-metaiodobenzylguanidine (MIBG) myocardial scintigraphy and/or polysomnographic confirmation of REM-sleep without atonia [2].
Early diagnosis of DLB can be challenging, especially in the absence of a caregiver, or when the caregiver does not live with the patient. In these situations, information about cognitive fluctuations, hallucinations and sleep behavior is often scarce. While parkinsonism can be initially absent.
Better detection of cognitive fluctuations during a consultation would be very helpful to diagnose DLB more easily and without the use of biomarkers, which are time and money consuming, not universally available and sometimes difficult to organize and obtain, especially in an ambulatory setting when there is no caregiver.
Few methods have been used to identify and quantify cognitive fluctuations without the help of caregiver questionnaires. Computerized assessment of variability of simple reaction time and choice reaction time is able to differentiate a group of DLB patients from AD patients and from normal controls, especially after a very short time frame (90 s) rather than after 1 h or 1 week [3]. This illustrates how impressive cognitive fluctuations in DLB can be and suggests that the time frame of a clinical consultation should be largely long enough to observe these fluctuations.
Bliwise and coworkers used intra-individual variations of a verbal forward and backward digit span test to analyze cognitive fluctuations by calculating a coefficient of variation (St dev/mean). They compared 13 patients with DLB with 64 Parkinson patients during 8 sessions over 48 h in a sound attenuated laboratory. A group difference was also seen with a more simple calculation, namely by subtracting for each patient the worst from the best score over the tested time [4].
We wanted to know if fluctuations in digit span test results could be used to differentiate DLB patients-individually and as a group-from patients with other dementias or from cognitively normal controls in a more ecological condition.

Methods
For this prospective study, patients were recruited from the Memory Clinics and the Geriatrics wards of our hospitals. Patients had to be able to understand and speak French and to give informed consent. Patients with delirium were excluded. They were divided in 3 groups: patients with a diagnosis of probable DLB (according to the fourth consensus criteria [2]), patients with a diagnosis of AD or vascular dementia, as defined respectively by the criteria of Dubois [5] and the NINDS-AIREN criteria [6], and a control group of ambulatory patients who had no history of cognitive decline. Demographical data such as age and gender were collected from the patient's identity card, the educational level was obtained from the patient.
The digit span is a subtest of the Wechsler Adult Intelligence Scale, Fourth Edition (WAIS-IV). In its forward condition (FDS), subjects have to repeat strings of digits of increasing length in the same order as presented, until they fail twice on items of the same length. In the backward condition (BDS), strings have to be repeated in reversed order [7]. The FDS task is associated with attentional capacity, where the BDS is more related to global executive functioning and working memory [8].
We used as scores the longest spans that could be reproduced for FDS and BDS.
The test consisted of two FDS and BDS trials with an interval of 20 min, in the morning or early afternoon for hospitalized patients and at the beginning and the end of a consultation for ambulatory patients.

Results
Sixty-five patients were included. Their demographical data can be found in Table 1. Of the 22 patients with other forms of dementia than DLB, 14 had AD and 8 had vascular dementia. Patients with DLB were significantly younger than those of other groups (p = 0.02), while controls had higher MMSE scores (p < 0.0001). Patients with other dementias had lower education than both other groups.
In the DLB group, parasomnia was the most frequently observed core criterion (78%), followed by visual hallucinations (65%), cognitive fluctuation (48%) and extrapyramidal signs (43%). Ten patients had undergone a ioflupane scan, which was abnormal in 80% of cases. Table 2 shows the mean scores for the FDS and BDS for all groups. There were no significant intergroup differences for raw FDS scores. For the BDS, controls did better than patients with other dementias on both trials (p = 0.003 and 0.001), but not better than patients with DLB.

Group comparison
Group results for FDS and BDS can be found in Table 2. Compared to controls and patients with other dementias, the DLB group showed significant performance fluctuations between both trials of the FDS (FDS-D, p = 0.0006) and of the BDS (BDS-D, p = 0.003). Seventy percent of patients with DLB showed evidence of cognitive fluctuations for at least one test, while only 9.5% of controls and patients with other dementias did (Fig. 1). The presence of cognitive fluctuation was not correlated with educational level: respectively 22% and 14% of the patients with and without cognitive fluctuation had not finished secondary education.

Discussion
Cognitive fluctuation, here defined as a different score after 20 min on either a forward or a backward digit span test, was present in 70% of patients with DLB and only in 9.5% of patients with other dementias or without cognitive problems. This is in line with the literature, which cites a prevalence in DLB up to 90% [9]. With a specificity for DLB of 90%, our repeated digit span test performed better than the Dementia Cognitive Fluctuation Scale (specificity 74-79%) and almost as well as the Clinician Assessment of Fluctuation Scale (which showed a specificity of 91% against anatomo-pathological confirmation) and the One Day Fluctuation Scale (95%)    [3,10]. It had a positive predictive value that was slightly lower than that of the Mayo Fluctuations Composite Scale (80 vs 83%) but had a better negative predictive value (84% vs 70%) [11]. A direct comparison would be useful in a cohort where all patients have a caregiver who lives with the patient. In our study, 73% of patients with anamnestic evidence of cognitive fluctuation showed cognitive fluctuation on the digit span test. Walker and coworkers used computerized reaction time tests over a 90 s interval to establish the presence of cognitive fluctuation. They obtained a specificity of 98% in discriminating DLB and AD cases, but sensitivity was much lower than in our study (46 vs 70%) [3].
The major flaw of this study is the ill-defined control group where some subjects had MMSE scores as low as 24.
We have chosen not to use a cut-off for MMSE to define the control group by fear of excluding patients with low levels of formal education. It is not impossible that some of the controls had cognitive problems but if this were the case then it would rather underestimate the test's validity. Low education was not associated with cognitive fluctuation, which suggests that our test is "culture-fair", meaning it can be used in different cultural settings. Other flaws are the relatively small size of the three groups, with significant variability in age and education levels between them.
Although test reproducibility over time and inter-rater reliability need yet to be studied, these findings suggest that a changing score on a repeated forward and backward digit span test is a reliable and easy bedside tool for detecting cognitive fluctuations whenever DLB is suspected, without the need for a caregiver or an experienced clinician. Moreover, the test proved to be easily understood by patients of different educational levels and took less than 5 min, when not considering the 20 min interval between the two sessions.
Author contributions KS and FB designed the protocol. The tests were performed by FOR, KS and FB. Literature research and statistical analysis and was done by FOR and KS. All authors contributed to the redaction of the article.
Data availability Raw data available on request.