Results

Demographics and clinical characteristics

The three patient groups were matched for age (all p values = 1.00; See Table A.1), years of education (F(3,298) = 1.88; p =0.134) and disease duration (e.g., years since symptom onset; H(2) = 4.95; p = 0.084). While the SD and control groups had a relatively equal distribution of males and females, the bvFTD groups showed clear male dominance (see Table 1). Disease severity (FRS), ranged from very mild to severe, with most of the patients in the moderate range. Disease severity in both bvFTD groups was greater than in the SD group (mean difference with probable bvFTD= − 1.77, 95% CI [− 2.42, − 1.13], p < 0.001; mean difference with possible bvFTD = − 1.40, 95% CI [− 2.14, − 0.67], p <0.001), but no significant difference was found between the two bvFTD groups (mean difference probable vs possible bvFTD = − 0.368, 95% CI [− 0.89, − 0.15], p = 0.212). On cognitive measure of disease severity, the reverse was found, where the SD group performed significantly below both bvFTD groups (total ACE-III, both p values ≤ 0.001; see Table A.2). Here, possible bvFTD outperformed probable bvFTD (z = − 3.22; p = 0.008).

 | Possible bvFTD(n = 43) | Probable bvFTD(n = 106) | SD(n = 42) | Controls(n = 112) | Group effect (F or H or χ2) | p value | Post hoc test results
Sex (m:f) | 36:7 | 68:38 | 24:18 | 55:57 | 16.56a | <0.001 | Male dominance in bvFTD
Age (years) | 62.20 (8.65) | 62.56 (8.07) | 64.58 (7.07) | 66.40 (6.89) | 5.79b | <0.001 | Both bvFTD < Controls
Education (years) | 12.61 (3.35) | 12.47 (3.09) | 12.48 (2.90) | 13.33 (2.69) | 1.88b | 0.134 | Nil
Disease duration (years) | 5.74 (3.60) | 4.45 (2.98) | 4.06 (2.07) | – | 4.950c | 0.084 | SD < Possible bvFTD
FRS | 0.02 (1.13) | − 0.35 (1.11) | 1.42 (1.51) | – | 29.87b | < 0.001 | Both bvFTD < SD
ACE-III Total (/100) | 85.19 (9.85) | 77.61 (10.43) | 61.38 (12.55) | 94.87 (3.11) | 201.19c | < 0.001 | SD< Probable bvFTD < Possible bvFTD; All patient groups < Controls

Cognitive profile

Patient groups performed significantly below healthy controls across all measures, with the exception of the ACE-III Visuospatial subscale and the ratio of TrailsB/TrailsA, where only the probable bvFTD group was reduced (ACE-III Visuospatial: z = − 6.03, p <0.001; ratio TrailsB/TrailsA: z =2.92; p =0.021; see Table 2 and Appendix A.2 and Appendix A.3). Among the patient groups, no significant differences were present in attentional capacity (all p values ≥ 0.22) or processing speed (all p values ≥ 0.372; see Appendix A.3). Some differences arose between groups in language, memory and executive function. The probable bvFTD group showed poorer working memory (DSB; p = 0.004) and slower performance on TMT-B (p = 0.009) than the SD group. The SD group was more impaired on the ACE-III language and memory subscales than both bvFTD groups (both p values < 0.001; Appendix A.2), and more impaired than the possible bvFTD group on ACE-III Fluency (p < 0.001; Appendix A.2).

 | Possible bvFTD(n = 43) | Probable bvFTD(n = 106) | SD(n = 42) | Controls(n = 112) | Group effect Kruskal–Wallis | p-values | Post hoc test (Bonferroni corrected)
ACE-III
Attention (/18) | 17.00 (2.00) | 16.00 (2.00) | 15.40 (1.85) | 17.40 (0.40) | 74.71 | < 0.001 | All patient groups < Controls
Memory (/26) | 22.00 (6.50) | 19.00 (6.00) | 12.00 (6.75) | 25.00 (2.25) | 158.71 | < 0.001 | SD < Probable bvFTD < Possible bvFTD < Controls
Fluency (/14) | 10.00 (6.50) | 7.00 (5.00) | 5.00 (5.00) | 13.00 (1.25) | 162.17 | < 0.001 | SD, Probable bvFTD < Possible bvFTD < Controls
Language (/26) | 25.00 (2.00) | 23.00 (4.00) | 12.50 (5.85) | 25.90 (1.00) | 152.89 | < 0.001 | SD< both bvFTD < Controls
Visuospatial (/16) | 15.00 (2.00) | 15.00 (2.00) | 16.00 (1.00) | 16.00 (1.00) | 38.33 | < 0.001 | Probable bvFTD < Controls
Digit span
Forward (max span) | 6.00 (2.00) | 6.00 (2.00) | 6.50 (2.75) | 7.00 (2.00) | 43.95 | < 0.001 | All patient groups < Controls
Backward (max span) | 4.00 (2.00) | 4.00 (2.00) | 5.00 (1.00) | 5.00 (2.00) | 78.23 | < 0.001 | Probable bvFTD < SD; All patient groups < Controls
Trail Making Test
Part A (sec) | 39.00 (17.00) | 45.50 (34.75) | 39.00 (22.00) | 30.00 (10.00) | 63.78 | < 0.001 | All patient groups < Controls
Part B (sec) | 95.00 (104) | 154.00 (90.25) | 85.50 (90.25) | 72.00 (33.00) | 88.12 | < 0.001 | Probable bvFTD < SD; All patient groups < Controls
Ratio (TrailsB/TrailsA) | 2.47 (1.11) | 2.65 (1.34) | 2.54 (1.26) | 2.26 (0.77) | 9.14 | 0.027 | Probable bvFTD < Controls
Verbal fluency
Letter fluency (F,A,S) | 31.00 (23.5) | 22.00 (19.00) | 25.00 (14.00) | 48.00 (17.00) | 129.58 | < 0.001 | Probable bvFTD < Possible bvFTD < Controls
Hayling
Overall scaled score | 5.00 (3.00) | 3.00 (4.00) | – | 6.00 (1.00) | 93.86 | < 0.001 | Probable bvFTD < Possible bvFTD < Controls

Both bvFTD groups were comparable to each other on the attention, language and visuospatial subscales of the ACE-III (all p values >0.196; Appendix A.2). On verbal fluency and memory subscales of the ACE-III, and on other measures of verbal generativity (FAS;_ p_ = 0.008; Hayling, p =0.033), the possible bvFTD group outperformed the probable bvFTD group (all p values <0.041).

Language performance within bvFTD

Group differences on the SYDBAT tasks are shown in Fig. 1. In line with our first hypothesis, probable bvFTD demonstrated significantly greater difficulties than possible bvFTD; however, this difference was only clearly observed on the semantic association subtest (Z = − 3.43,_ p_ = 0.004). While a trend towards a significant difference was found on naming (Z = − 2.61 p = 0.055), similar levels of deficit arose between the two bvFTD groups on comprehension (Z = − 2.15,_ p_ = 0.188), and repetition (Z = 0.130,_ p_ = 1.000).

Fig. 1: SYDBAT language performance across groups (possible bvFTD, probable bvFTD, SD patients and healthy controls); P values are denoted as follows: * = < 0.05; ** = < 0.01; *** = < 0.001; **** = < 0.0001

Impairment frequencies varied across subtests and between bvFTD groups. In the possible bvFTD group, naming and comprehension deficits were observed in 28% and 26% of cases respectively), with around a fifth of patients showing impairments in semantic association (21%). The proportion of cases showing impaired word repetition was lower, at 14%. In the probable bvFTD group, approximately half of the patients showed impairments in naming (51%) and semantic association (49%). Deficits in comprehension were found in 41%. Again, the frequency of impairment for word repetition was lower, affecting 18% of patients. See Appendix A.6 for more detail.

Language performance in SD and bvFTD

As expected, the SD group performed significantly lower than both the possible and probable bvFTD groups on the naming, comprehension and semantic association subtests (all p values < 0.001; see Fig. 1 and Appendix A.4). No differences were observed among the patient groups for the repetition subtest.

Within the SD group, naming performance was significantly more impaired than their performance on the comprehension and semantic association subtests (both p values ≤ 0.009; see Appendix A.7). Contrary to prediction, this pattern was not found in either of the two bvFTD groups. While naming, comprehension, and semantic association performance (all p values < 0.001) were significantly lower than repetition in the probable bvFTD group, no difference in performance was present between naming and the other subtests (naming vs comprehension: p =0.839; naming vs semantic association: p =0.354). For possible bvFTD, no differences were observed among any of the subtest pairs (all p values ≥ 0.131; see Appendix A.6), indicating a similar level of performance across all four subtests for this group.

Semantic and executive associations with language deficits in bvFTD

To investigate the factors associated with language deficits in the two bvFTD groups, standard multiple regression analyses were conducted to assess the relative contributions of language (ACE-III Semantic) and executive functioning measures (DSB, FAS, TMT-B, Hayling Overall) to performance on each SYDBAT subtest. In the probable bvFTD group, significant models were obtained for SYDBAT naming (F (5, 84) = 7.42, p < 0.001; R__2 = 31%), comprehension (F (5, 84) = 4.98, p < 0.001; R__2= 23%), and semantic association tasks (F (5, 84) = 6.83, p < 0.001; R__2 = 29%). In each case, ACE-III Semantic emerged as a significant predictor of subtest performance. The only additional predictor that approached significance was the Hayling test when completing the semantic association task (see Table 3). No significant linear relationship was found between performance on the SYDBAT repetition task and any of the predictor variables (F (5, 84) = 1.19, p = 0.320, R__2 = 0.07).

Group | SYDBAT subscale | Variable | B | SE B | β | p | Sr2
Probable bvFTD | Naming | ACE-III Semantic | 0.47 | 0.13 | 0.33 | 0.001* | 0.10
 | Comprehension | ACE-III Semantic | 0.33 | 0.10 | 0.32 | 0.002* | 0.10
 | Semantic Association | ACE-III Semantic | 0.44 | 0.12 | 0.34 | 0.001* | 0.11
 |  | Hayling Overall | 0.41 | 0.21 | 0.21 | 0.056 | 0.03
Possible bvFTD | Naming | ACE-III Semantic | 1.84 | 0.29 | 0.86 | < 0.001* | 0.39
 | Comprehension | Hayling Overall | 0.48 | 0.20 | 0.43 | 0.021* | 0.09
 |  | ACE-III Semantic | 0.38 | 0.19 | 0.35 | 0.054 | 0.06
 | Semantic Association | ACE-III Semantic | 0.65 | 0.26 | 0.45 | 0.016* | 0.11
 | Repetition | ACE-III Semantic | 0.37 | 0.10 | 0.69 | < 0.001* | 0.25
 |  | Digits Backwards | 0.38 | 0.12 | 0.55 | 0.004* | 0.16
 |  | Hayling Overall | − 0.26 | 0.10 | − 0.48 | 0.012* | 0.12

For possible bvFTD, significant results were observed for all four SYDBAT tasks: naming (F (5, 36) = 13.21, p < 0.001; R__2 = 65%), comprehension (F (5, 36) = 5.31, p = <0.001; R__2 = 42%), semantic association (F (5, 36) = 4.85, p = 0.002; R__2=40%) and repetition (F (5, 35) = 5.06, p =0.001; R__2 = 42%). Here, the ACE-III Semantic variable predicted language scores for naming, repetition, and semantic association. Unlike in the probable bvFTD patients, only performance on the Hayling task was a significant predictor for the Comprehension subtest, with semantic performance producing a marginally significant result (p = 0.054; see Table 3).

For the repetition subtest in possible bvFTD, effect sizes were constrained by restricted variability, given individual performances clustered near ceiling. Performance was predicted by the ACE-III Semantic variable and two measures of executive function: working memory (DSB), which provided a positive contribution, and response generation and inhibition (Hayling Overall), which showed a significant negative standardized coefficient (β = −0.48, p = 0.012). Given a negligible zero-order correlation between the Hayling and repetition performance (r = 0.11), this result suggested a suppression effect rather than a direct inverse association (i.e., indicating that performance on the Hayling may have accounted for variance in other measures unrelated to repetition performance). Collinearity diagnostics indicated acceptable tolerance and variance inflation factors (all VIFs < 2.0), suggesting that the observed suppression effect was not attributable to multicollinearity.