Quantitative and qualitative features of executive dysfunction in frontotemporal and Alzheimer’s dementia

Abstract

Behavioral variant fronto-temporal degeneration (bvFTD) is typically distinguished from Alzheimer’s disease (AD) by early, prominent dysexecutive findings, in addition to other clinical features. However, differences in executive functioning between these groups are not consistently found. The current study sought to investigate quantitative and qualitative differences in executive functioning between those with bvFTD and AD in a large sample, while controlling for dementia severity and demographic variables. Secondary data analyses were completed on a subset of cases from the National Alzheimer’s Coordinating Center collected from 36 Alzheimer’s Disease Research Centers and consisting of 1,577 individuals with AD and 406 individuals with bvFTD. Groups were compared on 1) ability to complete three commonly administered executive tasks (letter fluency, Trail Making Test Part B [TMTB], and digits backward); 2) quantitative test performance; and 3) errors on these tasks. Findings suggested that individuals with bvFTD were less likely to complete letter fluency, χ²(2) = 178.62, p < .001, and number span tasks, χ²(1) = 11.49, p < .001), whereas individuals with AD were less likely to complete TMTB, χ²(2) = 460.38, p < .001. Individuals with bvFTD performed more poorly on letter fluency, F(1) = 28.06, p = .013, but there were not group differences in TMTB lines per second or number span backwards. Errors generally did not differentiate the diagnostic groups. In summary, there is substantial overlap in executive dysfunction between those with bvFTD and AD, though individuals with bvFTD tend to demonstrate worse letter fluency performance.

Cognitive differences between bvFTD and AD

Accurate characterization of cognitive impairment is frequently required to help determine the etiology of and treatment plan for individuals with neurodegenerative conditions. To this end, Alzheimer’s disease (AD) and behavioral variant frontotemporal (bvFTD) dementias are thought to have different profiles of cognitive impairment early in the disease course. Clinical consensus criteria hold that AD is typically characterized early by an amnestic presentation (McKhann et al., 2011). In contrast, early and prominent deficits in executive and behavioral functioning are among the cardinal diagnostic features of bvFTD (Rascovsky et al., 2011).

Research support for this distinction rests on a broad array of findings supportive of a fundamental distinction between AD and bvFTD in terms of memory. Poos, Jiskoot, Papma, van Swieten, and van den Berg (2018) recently conducted a meta-analysis on this topic that included data from 20 studies comparing the memory performance of individuals with AD to those with bvFTD. AD patients performed significantly worse across all aspects of memory relative to the bvFTD cohort. Effect sizes were large for learning and recall and moderate for recognition when comparing the two conditions, although individuals with bvFTD still demonstrated significantly impaired memory relative to controls.

In contrast, the distinction between these conditions with regard to executive functioning (EF) is less clear. Before discussing the studies in more detail, it is important to note that EF itself is a multifaceted phenomenon without a readily agreed upon definition. EF is generally thought of as consistent of higher order thinking abilities, such as working memory, set shifting, planning, organization, problem-solving, and inhibition, which are mediated by frontal and frontosubcortical circuits (Heilman & Valenstein, 2010; Scott & Schoenberg, 2011). Due to its multifaceted nature, executive functioning is indexed by a host of different measures (Lezak, Howieson, & Loring, 2012). Thus, at least some of the heterogeneity in findings can be attributed to differences across studies in defining and measuring EF.

While definitions and conceptualizations of EF vary, neuropsychology has generally coalesced around common measures in clinical practice (Kudlicka, Clare, & Hindle, 2011). Pertinent to the current study, phonemic verbal fluency (i.e., generating as many words as possible that begin with a certain letter within a specified time limit) is typically considered a measure of executive functioning (Lezak et al., 2012) due to the need to perform “an unpracticted test that depends on effective strategies” of word retrieval (p. 693). These tasks are highly sensitive to frontal lobe dysfunction and likely engage the dorsolateral prefrontal cortex (Henry & Crawford, 2004). In a similar vein, rapid set shifting, as indexed by measures like Trail Making Test Part B (TMTB; Partington & Leiter, 1949), “is often used to assess executive functioning because of the contribution of mental flexibility when alternating between number and letter sets” (Lezak et al., 2012, p. 423). Performance on this measure has also been related to the frontal lobes from converging evidence using lesion-mapping and functional imaging paradigms (Varjacic, Mantini, Demeyere, & Gillebert, 2018). Finally, working memory tasks are often included in consensus batteries of executive functioning (Kramer et al., 2014). Working memory is thought to involve a so-called “central executive” (Baddeley, 1992; Baddeley & Della Sala, 1996) that allows for information held in “temporary storage to be manipulated for complex cognitive operations ... [and] is recruited to focus attention and combat interference” (Lezak et al., 2012, p. 408). One of the most common of these types of tests requires patients to repeat a series of digits in reverse order (Baddeley, 1992; Faria, Alves, & Charchat-Fichman, 2015). Performance on this task is consistently correlated with volume of the prefrontal cortex and recruitment of fronto-temporal networks (Li, Qin, Zhang, Jiang, & Yu, 2012; Yuan & Raz, 2014).

Studies specifically assessing EF differences between AD and bvFTD have yielded at least moderate support the traditional conceptualization of bvFTD as characterized by more dysexecutive features. For example, Libon et al. (2007) compared 16 bvFTD patients to 38 AD patients (diagnoses based on clinical consensus criteria) on a factor score consisting of an inhibition and a set switching task. They found a significant, small-to-moderate difference between the groups on this index favoring individuals with AD. In a similar, slightly larger study, Thompson, Stopford, Snowden, and Neary (2005) found that individuals with bvFTD performed worse on a letter fluency task than individuals with AD. Other studies have replicated this phonemic fluency finding, but did not find evidence of weaker performance among individuals with bvFTD on other executive indices involving working memory and set shifting (Possin et al., 2013; Wicklund, Rademaker, Johnson, Weitner, & Weintraub, 2007). Finally, Ritter, Leger, Miller, and Banks (2017) reported that individuals with bvFTD perform more poorly on digit span backward, a working memory measure, though they failed to find group differences on a digit symbol task and also reported that individuals with AD performed more poorly than those with bvFTD on a measure of set shifting.

In contrast, a number of investigations have failed to find significant differences between the two dementia groups on performance based executive measures. For instance, Wicklund, Johnson, and Weintraub (2004) reported that individuals with clinically diagnosed bvFTD did not differ from individuals with AD on a measure of reasoning. Similarly, Collette et al. (2007) found that the groups did not differ on a test of inhibition. There were no significant differences between bvFTD and AD on a tower test, a measure of conceptualization and problem-solving (Carey et al., 2008). Additionally, individuals with bvFTD did not differ from individuals from individuals with AD on a design fluency task in a study from Possin et al. (2012). Participants from the two dementia groups also do not appear to differ in overall clock drawing performance, a measure of planning and organization (Barrows et al., 2015).

Tellingly, Hutchinson and Mathias (2007) conducted a comprehensive meta-analysis of 93 studies examining differences in neuropsychological test performance between those with AD and any individuals with FTD spectrum disorders. Their analyses revealed effect sizes (Cohen’s d values) ranging from −0.60 (in favor of bvFTD) to 0.66 (in favor of AD), depending on the executive measure chosen. The authors noted that “... there was between 57% and 92% overlap in the scores of the AD and FTD groups on all 12 [executive] measures” (Hutchinson & Mathias, 2007, p. 922). Overall then, the distinction between these two groups in terms of executive performance is far from certain.

Beyond basic performance based scores, executive dysfunction can be defined by qualitative indices, such as errors produced during the completion of neuropsychological tasks. Thus, it may be that errors distinguish the groups. Here too, differences in executive phenomenology between AD and bvFTD are unclear. For example, Carey et al. (2008) reported that individuals with bvFTD committed more errors on a tower test than did individuals with AD. Similarly, in another study, bvFTD patients were more prone to repetition errors than AD patients on a design fluency task (Possin et al., 2012). Finally, bvFTD may be distinguished from AD by a higher likelihood of inaccurate hand placement on a clock drawing task (Barrows et al., 2015). However, other investigations have reported no differences between the groups in errors on letter fluency and inhibition tasks (Collette et al., 2007; Thompson et al., 2005).

Implications and limitations of current studies

The mixed findings and other evidence have led some researchers to call for rather significant reconceptualizations of the defining cognitive features of AD and bvFTD. For example, some authors have questioned the appropriateness of utilizing executive deficits as a cardinal feature of bvFTD (Dodich et al., 2017). Others have proposed that executive deficits in bvFTD are actually not an early marker of the disease, and rather represent a later stage (Roca et al., 2013). Additionally, there continues to be debate as to whether executive deficits emerge early or late in the progression of AD (Karantzoulis & Galvin, 2011). Furthermore, investigators have proposed that there may be a frontal variant of AD with both memory and executive impairments, which could create a confound when attempting to differentiate bvFTD from AD (Blennerhassett, Lillo, Halliday, Hodges, & Kril, 2014; Duker et al., 2012; Li, Zhou, Lu, Wang, & Zhang, 2016; Ryan et al., 2017; Wong et al., 2016; Woodward et al., 2009, 2010).

While these research findings could lead to a call for a reworking of clinical criteria, there are several other factors that deserve consideration in interpreting research on bvFTD and AD that may have contributed to equivocal findings. First, the samples utilized in previous studies tended to be small, such that minor differences in sample composition could produce different patterns of results across studies. For example, in the Hutchinson and Mathias (2007) meta-analysis previously discussed, the mean numbers of AD and bvFTD participants per study were 31.2 and 18.6, respectively.

Another possible reason for mixed findings concerns lack of control for global cognitive decline/dementia severity in comparisons of individuals with bvFTD and AD. Indeed, previous research suggests individuals with bvFTD decline significantly more quickly than those with AD on broad cognitive screening measures and indices of daily functioning (Rascovsky et al., 2005; Wicklund et al., 2007). Thus, matching groups without accounting for differences in dementia severity introduces a confound; that is, differences in EF performance between bvFTD and AD may be introduced or masked due to group differences in overall dementia severity.

Even when this possibility was controlled for in previous work, the method of control may have been inadequate. A common methodology in the aforementioned studies was to equate dementia severity by means of utilizing a global cognitive screening measure, most commonly the Mini Mental State Exam (MMSE; Cockrell & Folstein, 2002). Unfortunately, the MMSE tends to over sample from memory and language domains, which may be less impacted early in bvFTD (Nieuwenhuis-Mark, 2010). Thus, equating individuals from these two groups on the MMSE may not mean translate to an equivalent level of dementia severity.

Similarly, the influence of demographic variables on test performances has been inconsistently accounted for in prior studies. Individuals with bvFTD and AD differ on these variables, with bvFTD having a younger age of onset than AD (Roberson et al., 2005). Additionally, AD tends to be more common in females, whereas bvFTD is more common in males (Podcasy & Epperson, 2016). These demographic factors, and others like education, can impact performance on cognitive/executive measures even in the absence of pathology (Weintraub et al., 2018). Thus, lack of control for demographic variables is a serious limitation of previous studies.

There has been one large study that assessed differences in demographically corrected EF performance between bvFTD and AD across levels of dementia severity (Ranasinghe et al., 2016), using the Clinical Dementia Rating (CDR; Morris, 1993). Findings were complex. First, patients with bvFTD performed better than those with AD on digits backward, but only at the moderate stage of dementia severity. Second, individuals with AD performed worse than those with bvFTD on TMTB time, but only among those with very mild and mild dementia. Third, individuals with bvFTD performed worse than those with AD on TMTB errors and on a letter fluency task across all dementia severity levels.

These findings certainly warrant replication in a new sample. Additionally, the study included some important limitations that may be improved upon in follow up studies. First, Ranasinghe et al. (2016) chose an age-matched sample of individuals with AD for their research. This choice does not reflect the clinical reality that AD patients tend to be older than those with bvFTD, as previously discussed. Thus, it may have been more appropriate to include all individuals with AD and then correct for age using demographically adjusted scaled scores.

Second, the authors standardized neuropsychological test scores based on performance by a relatively small sample of healthy control participants (N = 126), as opposed to using a larger, well-developed standardization sample that mirrors the type of normative data typically used in clinical practice.

Third, the article by Ranasinghe et al. (2016), like many neuropsychological studies, did not report data on whether participants were able to actively participate in testing. Lack of ability to complete testing caused Ranasinghe and colleagues to exclude individuals with more severe dementia from their analyses, such that they did not capture the full range of dementia severity. This limitation raises the possibility that detection of differences between AD and bvFTD were confounded by a selection effect for individuals able to complete the measure in question. As the authors point out in their discussion, “Poor test performance of patients with bvFTD is not always attributable to cognitive deficits but rather to noncooperation with the testing procedures” (Ranasinghe et al., 2016, p. 608), a possibility that deserves empirical examination. In a rare report of such data, Wong et al. (2016) found that 28.57% of individuals with AD and 18% of individuals with bvFTD were either unable to complete TMTB or failed to do so within the time limit. Thus, inability to complete executive tasks may be another data point that distinguishes between dementia groups. Alternatively, the presence of missing data from patient cohorts could skew results and needs to be reported.

Fourth, Ranasinghe and colleagues (2016) reported results from TMTB without adjusting for differences in processing speed, which can drive poor performance on this task independent of set switching problems (Lamberty, Putnam, Chatel, & Bieliauskas, 1994; Sanchez-Cubillo et al., 2009; Senior, Piovesana, & Beaumont, 2018; Weintraub et al., 2018). Failure to correct for processing speed is particularly notable when studying dementia patients because deficits in graphomotor speed tend to drive poor performance on TMTB to a greater extent in cognitively impaired than cognitively intact populations (Misdraji & Gass, 2010).

Fifth, the authors only indexed errors from TMTB, failing to report on other important error metrics, such as those derived from phonemic fluency tests.

Current study

Thus, the goal of the current study was to replicate the findings of (Ranasinghe et al., 2016), while addressing the aforementioned limitations. Specifically, we examined the quantitative and qualitative characteristics of executive dysfunction in individuals with bvFTD and AD and identified in what stages of dementia severity these distinctions occur. We addressed several specific hypotheses:

1. We present data on whether or not individuals in the two dementia groups were able to complete executive tasks at each level of dementia severity. Per the above argument put forth by Ranasinghe and colleagues (2016), we expected that individuals with bvFTD would be less likely than those with AD to engage in testing.

2. We examined quantitative differences in executive performance between individuals with bvFTD and AD across levels of dementia severity. We expected to replicate the findings of Ranasinghe et al. (2016):

   1. Slower TMTB performance for AD at the very mild and mild dementia stages We further hypothesized that these group differences for TMTB would fail to reach significance after adjusting for processing speed, as TMTB impairment is often better accounted for by processing speed deficits (Misdraji & Gass, 2010).
   2. Better performance across all levels of dementia severity for AD on phonemic fluency.
   3. Better performance on number span backwards for individuals with bvFTD at the moderate-severe stage of dementia.

3. We evaluated errors across bvFTD and AD within each level of dementia severity. We expected that individuals with AD would be more error prone on TMTB than those with bvFTD, similar to the findings of Ranasinghe et al. (2016). However, we hypothesized that errors on phonemic fluency would be more common in bvFTD, going along with traditional findings of poorer overall phonemic fluency performance in this group (Possin et al., 2013; Ranasinghe et al., 2016; Thompson et al., 2005; Wicklund et al., 2007).

Additional steps were taken to overcome limitations of previous studies, including use of a large sample, diagnosis of bvFTD and AD using clinical consensus criteria, and measurement of executive functioning with widely accepted instruments with broad normative data and appropriate demographic corrections (Devora, Beevers, Kiselica, & Benge, 2019; Weintraub et al., 2018). Findings from this research are important to improve the overall conceptualization and differential diagnosis of bvFTD and AD, assist in treatment planning, and guide clinical trial design.

Methods

Sample

Analyses were carried out on a subset of cases from the National Alzheimer’s Coordinating Center (NACC) Uniform Dataset (UDS). The initial data request included all participants with baseline UDS data recruited from 36 Alzheimer’s disease Research Centers (ADCs) from 2015 to February, 2018. This sample was refined to include only individuals classified as having at least at least some impairment on the Clinical Dementia Rating global score (i.e., 0.5 or higher). Next, the dataset was reduced to only those individuals diagnosed as having AD (N = 1577) or bvFTD (N = 406) as their primary etiology. Diagnoses were made by assessing clinicians or teams at each ADC by reference to the most up-to-date clinical consensus criteria available at the time of diagnosis (Morris et al., 2006; Weintraub et al., 2018); that is, published criteria from Rascovsky et al. (2011) were used for bvFTD, while consensus standards by McKhann et al. (2011) were used for AD.

Measures and procedures

All participants and collaterals completed measures of cognitive and day-to-day functioning, in addition to providing demographic information, in accordance with procedures established by the NACC (ADC Clinical Task Force, & National Alzheimer’s Coordinating Center, 2015). Data are collected under the auspices of the internal review boards at the individual ADCs and made available to qualified researchers via an online request portal. All data were provided in a deidentified format, and ethical guidelines for data use were followed throughout the conduct of this research.

Measures utilized in the current study are described in the following section. All de-identified data was obtained from the NACC repository in a data request in March 2018. Study participants provided informed consent at their parent institution with all appropriate legal and review processes in place.

Dementia severity

Dementia severity was assessed via the CDR, a structured clinical interview for staging of dementia severity (Morris, 1993). The CDR exhibits good reliability and validity for dementia staging (Rikkert et al., 2011). Global scores are well established for this instrument, with a total score of 0 reflective of no impairment (benign cognitive difficulties with no deficits in self-care), 0.5 = questionable impairment (difficulties on some complex day-to-day tasks, but no deficits in self-care; very mild dementia), 1 = mild impairment (deficits in self-care for complex tasks, with preserved ability to complete basic activities of daily living; mild dementia), 2 = moderate impairment (assistance required for at least some basic activities, such as bathing, grooming, toileting, and eating; moderate dementia), and 3 = severe impairment (approaching or at complete dependence for basic day-to-day tasks; severe dementia). Given low base rates of severe impairment (CDR = 3) in the current sample, this group was combined with the moderate severity group (CDR = 2) and termed the moderate- severe dementia group. This grouping is often used in common parlance, as well as research (e.g., Galasko et al., 2005).

Quantitative indices of EF

EF was assessed via TMTB, letter fluency, and number span backward tests from the UDS 3.0 neuropsychological battery (Weintraub et al., 2018). These indices have excellent theoretical and research backing as measures of EF sensitive to frontal pathology (Baddeley, 1992; Baddeley & Della Sala, 1996; Henry & Crawford, 2004; Lezak et al., 2012; Li et al., 2012; Varjacic et al., 2018; Yuan & Raz, 2014). Additionally, they appear to be among the most commonly employed indices of EF in previous studies that have assessed differences in bvFTD and AD (Hutchinson & Mathias, 2007; Libon et al., 2007; Possin et al., 2013; Ranasinghe et al., 2016; Ritter et al., 2017; Thompson et al., 2005; Wicklund et al., 2007).

TMTB requires an individual to use a pen to engage in alternative, rapid sequencing of numbers and letters that have been printed in an array. The variable of interest is the time it takes the individual to correctly complete the array. This number is an estimate of rapid set switching skill, an executive ability, but is also affected by psychomotor speed and visual scanning abilities (Jarvis & Barth, 1994; Lezak et al., 2012; Strauss, Sherman, & Spreen, 2006). A related index is calculated as the number of lines per second that an individual completes on TMTB (Weintraub et al., 2018). It provides a slight correction for processing speed, in addition to allowing for inclusion of individuals who “time out” on the task; that is, those who reach the 300 second time limit without finishing the array. Both the total time and lines per second are utilized in the current evaluation.

Next, letter fluency tasks assess the number of unique words an individual can name that begin with a certain letter within one minute (in this case there were two trials for F and L). On the UDS, individuals are instructed to say any word beginning with the letter, but leave out numbers and names of people or places. Letter fluency involves working memory, cognitive flexibility, and word knowledge (Lezak et al., 2012; Shao, Janse, Visser, & Meyer, 2014).

Finally, number span backward requires individuals to hold a series of digits in short-term storage and then repeat them in reverse order. It is thought to be an indicator of working memory, mental manipulation, and inhibition (Baddeley, 1992; Baddeley & Della Sala, 1996; Lezak et al., 2012).

Qualitative observations

A previously understudied index of cognitive functioning on executive measures determines whether patients are willing/able to complete executive tasks. In the UDS, participants were coded as completers versus noncompleters on TMTB, letter fluency, and number span backward. Among noncompleters, the reason for not completing the test was recorded as a physical problem, cognitive/behavioral problem, other problem, or verbal refusal by the examining professional according to uniform procedures established by the ADC Clinical Task Force, & National Alzheimer’s Coordinating Center (2015).

Among individuals who did complete executive tasks, dysfunction was also indexed in the form of errors on these measures. Errors on TMT included any incorrect lines (i.e., commissions). Errors on letter fluency included repetition errors (repeating the same word twice) and rule violations (using a non-F or non-L word; using a non-word; using a number or name of a place or person).

Analyses

Data preparation

Descriptive statistics by dementia group for raw data were calculated. Next, all quantitative indices of EF were corrected for the influence of demographics, including age, sex, and education, using the UDS Norms Calculator (Weintraub et al., 2018) to convert raw scores into demographically-corrected z-scores. Cognitive variables were also transformed such that higher/more positive scaled scores indicated better performance. Analyses were carried out with missing cases removed list-wise in SPSS 24.0 (IBM, 2016).

Quantitative executive performance

Among those that completed the EF measures, we tested for differences in norm referenced executive task performance between individuals diagnosed as AD and bvFTD on TMTB time, TMTB lines per second, letter fluency overall score, and number span backwards total score. Specifically, we performed two-way Analyses of Variance with main effects of dementia group and dementia severity, as well as the interaction between these two variables. Effects were considered statistically significant when p < .05.

Qualitative executive performance

In the whole subsample, we also assessed the extent to which individuals with AD differed from those with bvFTD in their likelihood of completing TMTB, letter fluency, and number span backward tasks across levels of dementia severity via a chi-square test of independence (complete vs. non-complete X dementia group X dementia severity). When the overall test was significant (i.e., p < .05), follow up chi-square tests of independence were used to evaluate differences between AD and bvFTD within each severity group.

Finally, we tested for differences in errors on TMTB and letter fluency between AD and bvFTD groups across levels of dementia severity among those who had completed testing. Research by Devora et al. (2019) indicates that errors on these measures are low base rate phenomena with highly skewed distributions. To account for these factors, participants were categorized as having committed no errors or at least 1 error. Subsequently, a chi-square test of independence was used to examine differences between individuals with AD and bvFTD in the rate of having committed at least one error across levels of dementia severity. When the overall test was significant (i.e., p < .05), follow up chi-square tests of independence were used to evaluate differences between AD and bvFTD within each severity group.

Results

Demographic and descriptive data

Individuals with AD were 52.20% females and were primarily white (85.40%). They averaged 71.21 (SD = 9.17) years of age and completed on average 15.61 (SD = 3.02) years of education. The bvFTD sample included 39.70% females and 94.10% white individuals. The average age was 62.75 (SD = 8.22) and the mean years of education was 15.70 (SD = 2.97). In keeping with the trend for bvFTD to have an earlier age of onset than AD (Roberson et al., 2005), the bvFTD group tended to be younger, t(1981) = 16.91, p < .001. Additionally, in line with previous research on sex differences between AD and bvFTD (Podcasy & Epperson, 2016), the bvFTD group included fewer females, χ²(1) = 20.29, p < .001. The cohorts were similar in terms of education t(1954) = −0.52, p = .601.

The breakdown of participants into dementia severity groups among those with presumed AD was as follows: very mild (62.80%), mild (28.90%), and moderate-severe (8.30%). The breakdown for participants with presumed bvFTD included very mild (32.50%), mild (46.30%), and moderate-severe (21.20%). Dementia severity differed by dementia etiology, χ²((2) = 131.25, p < .001, such that individuals with AD were more likely to fall in the very mild group, whereas individuals with bvFTD were more likely to fall into the mild and moderate-severe groups.

Quantitative executive performance by dementia group and dementia severity

Descriptive statistics for raw cognitive variables can be found in Table 1. Comparisons between bvFTD and AD groups on these variables are found in Table 2. Results revealed a significant interaction of dementia group and dementia severity on TMTB time. Examination of this interaction revealed that while individuals with AD and bvFTD performed similarly on the task in the very mild group, individuals with AD tended to perform more poorly with worsening dementia severity.

Table 1.

Raw and demographically adjusted scaled score descriptive statistics by dementia group.

	AD (n = 1577)			bvFTD (n = 406)
	M (SD)	Median (IQ Range)	na	M (SD)	Median (IQ Range)	na
TMTB time raw	164.07 (91.19)	132.00 (87.00, 280.00)	1572	160.65 (88.63)	133.00 (87.00, 236.25)	400
TMTB time scaled	−1.97 (2.19)	−1.16 (−4.46, −0.17)	1135	−2.18 (2.13)	−1.45 (−3.85, −0.46)	274
TMTB lines by second raw	0.20 (0.17)	0.18 (0.09, 0.28)	1569	0.20 (0.14)	0.18 (0.10, 0.28)	387
TMTB lines by second scaled	−1.33 (1.29)	−1.38 (−2.05, −0.73)	1132	−1.68 (1.03)	−1.72 (−2.43, −1.09)	261
TMTB errors	1.64 (2.64)	1.00 (0.00, 2.00)	1569	1.98 (2.65)	1.00 (0.00, 3.00)	387
Letter fluency total words raw	20.78 (9.50)	20.00 (14.00, 27.00)	1472	14.39 (9.80)	13.00 (6.00, 20.25)	129
Letter fluency total words scaled	−0.94 (1.14)	−0.98 (−1.75, −0.17)	1384	−1.84 (1.22)	−1.94 (−2.86, −1.02)	113
Letter fluency repetition errors	1.78 (2.08)	1.00 (0.00, 3.00)	1472	1.13 (1.57)	1.00 (0.00, 2.00)	129
Letter fluency rule violations	1.02 (1.76)	0.00 (0.00, 1.00)	1472	0.89 (1.60)	0.00 (0.00, 1.00)	129
Number span backwards raw	4.82 (2.34)	5.00 (3.00, 6.00)	1472	4.48 (2.68)	4.00 (3.00, 6.00)	129
Number span backwards scaled	−1.03 (1.08)	−1.01 (−1.77, −0.27)	1393	−1.40 (1.21)	−1.49 (−2.40, −0.57)	117

Note. AD = Alzheimer’s disease; bvFTD = behavioral variant fronto-temporal degeneration; TMTB = Trail Making Test Part B.

^aReported n values for raw scores are for the number of individuals to whom the test was attempted to be administered. Thus, they exclude individuals with purely missing data, but include those individuals without a value on the task due to attempt but non-completion. Scaled score n values include only completers with available demographic data for correction.

Table 2.

ANOVA results for executive performance among completers.

	M (SD)			F, p
	AD	bvFTD	Both groups	Main effect of dementia severity	Main effect of dementia group	Dementia severity by dementia group interaction
TMTB time				16.33, .058	3.04, .175	3.73, .024
scaled score
Very mild	−1.57 (2.00)	−1.52 (1.83)	−1.57 (1.98)
Mild	−3.14 (2.29)	−2.32 (2.08)	−2.86 (2.25)
Moderate-severe	−4.97 (1.43)	−4.02 (2.21)	−4.39 (1.98)
All severities	−1.98 (2.19)	−2.18 (2.13)	−2.01 (2.18)
TMTB lines by second				38.72, .025	0.00, .956	0.79, 456
scaled score
Very mild	−1.17 (1.32)	−1.36 (1.02)	−1.19 (1.29)
Mild	−1.76 (1.04)	−1.80 (0.98)	−1.77 (1.01)
Moderate-severe	−2.61 (0.79)	−2.38 (−0.87)	−2.47 (0.84)
All severities	−1.33 (1.29)	−1.68 (1.03)	−1.39 (1.25)
Letter fluency				40.77, .024	28.06, .013	0.90, .405
scaled score
Very mild	−0.72 (1.07)	−1.18 (1.24)	−0.73 (1.08)
Mild	−1.19 (1.12)	−1.99 (1.11)	−1.30 (1.16)
Moderate-severe	−2.07 (0.89)	−2.73 (0.81)	−2.17 (0.90)
All severities	−0.94 (1.14)	−1.85 (1.21)	−1.01 (1.17)
Number span				12.44, .074	0.63, .502	2.05, .129
backwards
scaled score
Very mild	−0.80 (0.98)	−1.23 (1.02)	−0.81 (0.98
Mild	−1.29 (1.09)	−1.28 (1.29)	−1.29 (1.12)
Moderate-severe	−2.14 (1.01)	−2.10 (1.03)	−2.13 (1.01)
All severities	−1.03 (1.08)	−1.40 (1.21)	−1.05 (1.10)

Note. ANOVA = analysis of variance; AD = Alzheimer’s disease; bvFTD = behavioral variant fronto-temporal degeneration; TMTB = Trail Making Test Part B; very mild = mild cognitive impairment.

p < .05.

There was no significant interaction effect when examining TMTB lines by second, however. Additionally, there was no main effect of dementia group for this index, though individuals with greater dementia severity did tend to perform more poorly.

Regarding letter fluency, there were main effects of dementia severity and dementia group. Individuals with greater disease severity performed more poorly on this task on average. Additionally, individuals with bvFTD tended to be have worse letter fluency performance. The interaction between dementia severity and dementia group was not statistically significant.

Finally, on number span backwards, there were no significant differences between the dementia groups. The impact of dementia severity and the dementia group by dementia severity interaction were also nonsignificant.

Qualitative executive performance by dementia group and dementia severity

Rates of completion vs. non-completion on TMTB and letter fluency are reported in Table 3. A chi-square test of independence assessing likelihood of test completion by dementia group and dementia severity revealed a significant overall difference for TMTB, χ²(2) = 460.38, p < .001. Follow up chi-squares within each level of dementia severity revealed that the AD and bvFTD groups did not differ in their likelihood of completing TMTB within the very mild group. However, individuals with AD were much less likely than individuals with bvFTD to complete TMTB in the mild and moderate-severe groups (see Table 3).

Table 3.

Rates of completion on executive function tasks by dementia group and dementia severity.

	TMTB			Letter Fluency			Number Span Backward
	n (%)			n (%)			n (%)
	AD	bvFTD	χ2,P	AD	bvFTD	χ2,p	AD	bvFTD	χ2,p
Very mild
Completers	879 (88.90%)	115 (87.80%)	0.14, .710	915 (98.10%)	35 (100.00%)	0.69, .407	912 (97.70%)	33 (94.30%)	1.75, .187
Non-completers	110 (11.10%)	16 (12.20%)		18 (1.90%)	0 (0.00%)		21 (2.30%)	2 (5.70%)
Physical problem	4 (0.30%)	0 (0.00%)		–	–		–	–
Cognitive/behavioral problem	79 (8.00%)	10 (7.60%)		3 (0.30%)	0 (0.00%)		2 (0.20%)	0 (0.00%)
Other problem	19 (1.90%)	3 (2.30%)		8 (0.90%)	–		15 (1.60%)	1 (1.50%)
Verbal refusal	8 (0.80%)	3 (2.30%)		9 (0.80%)	0 (0.00%)		4 (0.40%)	–
Mild
Completers	250 (55.20%)	135 (73.40%)	18.09, <.001	396 (95.0%)	63 (95.50%)	0.03, .865	401 (96.20%)	65 (98.50%)	.905, .342
Non-completers	203 (44.80%)	49 (26.60%)		21 (5.00%)	3 (4.50%)		16 (3.80%)	1 (1.50%)
Physical problem	5 (1.10%)	0 (0.00%)		–	–		–	–
Cognitive/behavioral Problem	168 (37.10%)	35 (19.00%)		8 (1.90%)	1 (1.50%)		6 (1.40%)	0 (0.00%)
Other problem	17 (3.80%)	11 (6.00%)		10 (2.40%)	–		8 (1.90%)	1 (1.50%)
Verbal refusal	13 (2.90%)	3 (1.60%)		3 (0.70%)	2 (3.00%)		2 (0.50%)	–
Moderate-Severe
Completers	19 (14.60%)	30 (35.30%)	12.49, <.001	92 (75.40%)	16 (57.10%)	3.77, .052	99 (81.10%)	20 (71.40%)	1.31, .252
Non-completers	111 (85.40%)	55 (64.70%)		30 (24.60%)	28 (42.90%)		23 (18.90%)	8 (28.60%)
Physical problem	1 (0.80%)	2 (2.40%)		–	–		–	–
Cognitive/behavioral Problem	100 (76.90%)	48 (56.50%)		22 (18.00%)	9 (32.10%)		21 (17.20%)	8 (28.60%)
Other problem	5 (3.80%)	3 (3.50%)		4 (3.30%)	–		0 (0.00%)	0 (0.00%)
Verbal refusal	5 (3.80%)	2 (2.40%)		4 (3.30%)	3 (10.70%)		2 (1.60%)	–

Note. TMTB = Trail Making Test Part B; very mild = mild cognitive impairment; AD = Alzheimer’s disease; bvFTD = behavioral variant fronto-temporal degeneration. Chi-square results are for rates of completion versus noncompletion across dementia groups within each level of severity.

p < .05.

A significant overall effect was also observed for letter fluency, χ²(2) = 178.62, p < .001. Follow up chi-squares revealed that while individuals with bvFTD were more likely to be non-completers than individuals with AD overall (11.60% vs. 4.70%, respectively; χ²(1) = 11.49, p < .001), there were no significant differences between dementia groups within each level of dementia severity (see Table 3).

Finally, a significant overall effect was observed for number span backwards χ²(2) = 103.89, p < .001. Follow up chi-squares revealed that individuals with bvFTD were more likely to be non-completers than individuals with AD overall (8.50% vs. 4.10%, respectively; χ²(1) = 5.45, p < .019); however, there were no significant differences between dementia groups within the different dementia severity classifications (see Table 3).

Descriptive data for the presence vs. absence of errors are presented in Table 4. The likelihood of committing a least one error on each executive test was assessed by dementia group and dementia severity via chi-square tests of independence. The overall test was not significant for TMTB commission errors, χ²(1) = 0.06, p = .812, or rule violations on letter fluency, χ²(1) = 1.02, p = .314. However, there was a significant overall effect for letter fluency repetition errors, χ²(1) = 11.24, p = .001. Follow up analyses within each dementia group revealed that individuals with AD were significantly more likely than individuals with bvFTD to commit repetition errors in the very mild stage (see Table 4). However, the groups were similarly likely to commit repetition errors in the mild and moderate-severe groups.

Table 4.

Frequency of committing at least one error on EF tasks by dementia group and dementia severity.

		n, (%)
	AD	bvFTD	Both groups	χ2, p
TMTB commission errors
Very mild	509 (58.00%)	60 (55.00%)	569 (57.70%)
Mild	183 (73.50%)	79 (61.20%)	262 (69.30%)
Moderate-severe	16 (84.20%)	24 (82.80%)	40 (83.30%)
All severities	708 (61.80%)	163 (61.00%)	871 (61.70%)
Letter fluency repetition errors
Very mild	621 (67.90%)	16 (2.50%)	637 (67.10%)	7.49, .006
Mild	268 (67.70%)	35 (44.60%)	303 (66.00%)	3.56, .059
Moderate-severe	54 (58.70%)	8 (50.00%)	62 (57.40%)	0.42, .516
All severities	943 (67.20%)	59 (51.80%)	1002 (66.10%)
Letter fluency rule violations
Very mild	405 (44.30%)	18 (51.40%)	423 (44.50%)
Mild	195 (49.20%)	23 (36.50%)	218 (47.50%)
Moderate-severe	47 (51.10%)	6 (37.50%)	53 (49.10%)
All severities	647 (46.10%)	47 (41.20%)	694 (45.70%)

Note. TMTB = Trail Making Test Part B; very mild = mild cognitive impairment; AD= Alzheimer’s disease; bvFTD = behavioral variant fronto-temporal degeneration. Chi-square results are for rates of individuals having completed 1 vs. 0 errors across dementia groups within each level of severity.

p < .05.

Discussion

The goal of this study was to compare quantitative and qualitative indices of executive dysfunction in individuals with AD and bvFTD, accounting for demographic influences and dementia severity. First, we examined how individuals with these conditions compare in their ability to complete the tests in question. Second, we tested differences across AD and bvFTD on traditional neuropsychological metrics of task performance. Finally, we assessed differences in the likelihood of committing errors on executive tasks across the two dementia etiologies. The results provide new insights into the differential diagnosis and staging of AD and bvFTD.

Failure to perform a task: An important datapoint?

To our knowledge, no studies have directly examined whether individuals with AD and bvFTD differ in their willingness/ability to engage in executive tests. Our findings were mixed in this regard. On the one hand (and contrary to our hypothesis), individuals with AD dementia were less likely to complete TMTB than those with bvFTD dementia (but this was not the case in the very mild group). This effect was fairly large, with an 18–19% higher completion rate in the bvFTD group. The explanation for this phenomenon is not clear, but one hypothesis concerns the impact of memory problems. Although both individuals with AD and those with bvFTD tend to have progression of memory problems over time, the memory of individuals with AD is much worse initially (Smits et al., 2015). Thus, with increasing disease severity, individuals with AD are at greater risk than those with bvFTD for experiencing impairing levels of memory dysfunction that may interfere with recall of task instructions. In our clinical experience, for example, many individuals with AD simply cannot remember the instructions on TMTB to complete the sample, such that we do not administer the task itself.

On the other hand, in accordance with expectations, individuals with bvFTD were approximately 7% and 4% less likely than individuals with AD to complete letter fluency and number span backward, respectively. These effects appeared to be driven by non-significant group differences in moderate-severe group, which raises the possibility that there are specific deficits at this stage that prevent performance of word generation and backward digit repetition tasks among those with bvFTD but not AD. For instance, in bvFTD the disease may spread to impact inferior frontal and lateral temporal lobes such that underlying aphasia prevents valid task performance in a manner not observed for the less linguistically mediated TMT measures. This possibility remains open to empirical examination.

From both clinical and research standpoints, the inability to complete a neuropsychological measure may still provide useful data. Clinically, the inability to perform such tasks may indicate the need to explore language comprehension, memory, or adaptive communication strategies to help with task performance day to day. From a research standpoint, more work is needed to help understand the real-world correlates of such behavior. Finally, from a statistical perspective, these results urge authors to accurately report the prevalence of non-completion and their attempts to account for this missing data, as it occurs often in these populations.

Quantitative executive performance

There were also mixed findings on dementia group differences for quantitative indices of executive dysfunction. First, we partially replicated findings from (Ranasinghe et al., 2016), who found that individuals with AD performed worse on TMTB at the very mild and mild dementia stages (Ritter et al., 2017, also reported superior performance among individuals with bvFTD, but did control for dementia severity). We also found that individuals with AD performed worse on TMTB, but only at the mild and moderate-severe dementia stages. Interestingly, there were no group differences on the TMTB lines per second score. This finding suggests that differences in TMTB performance between AD and bvFTD may be driven in part by reduced processing speed in the AD group. Regardless, these findings suggest that poor alternation of cognitive set, a component of executive dysfunction, is not uncommon in AD and is not a reliable differentiator between those with bvFTD and AD.

In contrast, letter fluency performance tended to be worse among those with bvFTD than AD. This finding mirrors prior studies (Possin et al., 2013; Ranasinghe et al., 2016; Thompson et al., 2005; Wicklund et al., 2007), suggesting it is robust. Indeed, the effect size for the group difference was large in this study (d = .77) and in previous investigations (among those that reported this information). Phonemic fluency impairments have been related in MRI studies to bilateral frontal lobe atrophy (Libon et al., 2009), which is typically more severe in early bvFTD than early AD and progresses more quickly (Krueger et al., 2010) in this condition. Thus, phonemic fluency deficits in particular may prove to be unique relative cognitive weaknesses in individuals with bvFTD. Further work is needed to see if interpreting phonemic fluency in relation to semantic fluency (Canning, Leach, Stuss, Ngo, & Black, 2004; Devora et al., 2019; Gladsjo et al., 1999; Strauss et al., 2006) provides additional unique separation between these two groups.

Finally, in contrast to Ranasinghe et al. (2016) who found some evidence of superior performance for bvFTD versus AD on the number span backward, we found no significant differences between the dementia groups on this task. Backward digit repetition requires proper functioning of networks involving the anterior temporal and frontal (i.e., anterior cingulate and insular cortices) regions (Li et al., 2012). In AD, pathology tends to spread from the medial temporal lobes to the lateral temporal and anterior parietal lobes, preserving the frontal and anterior temporal lobes until later in the disease process (Whitwell, 2010). Similarly, bvFTD tends to be characterized by prominent ventromedial atrophy, with anterior frontal and temporal lobe atrophy emerging later (Whitwell, 2010). Thus, dysfunction in the networks necessary to perform number span backward may emerge similarly across bvFTD and AD, failing to create a distinction between the two groups.

Qualitative executive performance

Regarding errors, we failed to replicate findings from (Ranasinghe et al., 2016), who reported more errors on TMTB for AD vs. bvFTD. We found that there were no significant group differences in likelihood of making at least 1 TMTB commission error or letter fluency rule violation. These findings suggest that there is not a general difference in error proneness between the groups. Results do indicate, however, that there may be a higher likelihood of verbal fluency repetition errors in the AD group at the very mild stage (in contrast to Thompson et al., 2005, who found no differences between groups but did not control for dementia severity). It may be that early memory problems among those with AD lead to a higher likelihood of repeating due to forgetting previously provided responses (Smits et al., 2015), and the source of early repetition errors in AD could be investigated in future research.

Limitations and future research

A limitation to the current work includes the cross-sectional nature of the design, which precludes definitive conclusions about the longitudinal progression of executive dysfunction in the two studied dementias. Future research could replicate our findings longitudinally. Additionally, we chose three common indices of executive functioning that have been shown to differentiate AD and bvFTD in previous research. As mentioned in the introduction though, executive abilities are quite varied, and further study of executive differences in AD and bvFTD with alternative metrics is warranted to provide a more comprehensive understanding of the topic. Next, there was a fair amount of missing data for some of the variables of interest in this study for the bvFTD sample. Even accounting for this missing data, though, the sample remains one of the largest analyses of executive functioning data among bvFTD patients reported to date. Another limitation of the UDS dataset is that it includes a fairly well educated and predominantly Caucasian sample, limiting the generalizability of findings to more diverse cultural and demographic backgrounds. Consequently, replication of this study in a low socioeconomic status and/or more diverse sample would be fruitful. Finally, further systematic study of qualitative features of task discontinuation may prove beneficial for understanding reasons behind differences detected in the current investigation.

Summary and conclusions

Our findings suggest that there is substantial overlap between AD and bvFTD in terms of executive functioning that can complicate differential diagnosis. Nonetheless, there are some subtle signs that may differentiate the groups. For example, individuals with AD may be less likely to complete TMTB. Additionally, individuals with bvFTD may demonstrate poorer letter fluency performance and tend to be more repetitive on this task early in the disease process. Future work will replicate these findings with an eye towards identifying neuroanatomical substrates of executive performance and offering explanations for the mechanisms of differential test findings between bvFTD and AD. These steps will ultimately improve differential diagnosis, enable more accurate prognostic statements, and inform clinical trial design.