Early within therapy naming probes as a clinically-feasible predictor of anomia treatment response

ABSTRACT

This study investigated the relationship between early within-therapy probe naming performance and anomia therapy outcomes in individuals with aphasia. Thirty-four adults with chronic, post-stroke aphasia participated in the Aphasia Language Impairment and Functioning Therapy (Aphasia LIFT) programme, comprised of 48 h of comprehensive aphasia therapy. Sets of 30 treated and 30 untreated items identified at baseline were probed during impairment therapy which targeted word retrieval using a combined semantic feature analysis and phonological component analysis approach. Multiple regression models were computed to determine the relationship between baseline language and demographic variables, early within-therapy probe naming performance (measured after 3 h of impairment therapy) and anomia treatment outcomes. Early within-therapy probe naming performance emerged as the strongest predictor of anomia therapy gains at post-therapy and at 1-month follow-up. These findings have important clinical implications, as they suggest that an individual’s performance after a brief period of anomia therapy may predict response to intervention. As such, early within-therapy probe naming may provide a quick and accessible tool for clinicians to identify potential response to anomia treatment.

KEYWORDS:

• Aphasia
• language
• anomia
• rehabilitation
• behavioural predictors

Introduction

Aphasia is a debilitating communication impairment frequently resulting from stroke (Engelter et al., 2006) that can negatively affect individuals’ ability to convey thoughts and emotions, engage in conversation and participate in daily communicative activities (Best et al., 2011; Hilari, 2011; Ledorze & Brassard, 1995; Maher & Raymer, 2004). Anomia is recognized as a predominant feature of aphasia and as such, much research has been directed to its remediation (for a review see Laine & Martin, 2006; Nickels, 2002b; Whitworth et al., 2005). While there is evidence supporting the general benefits of anomia therapy, individuals’ response to therapy remains variable, and it is difficult to predict who will respond to therapy (Best & Nickels, 2000; Charidimou et al., 2014; REhabilitation and recovery of peopLE with Aphasia after StrokE (RELEASE) Collaborators, 2021).

Studies investigating recovery of general language functions in aphasia indicate that spontaneous and treatment-induced recovery is multifactorial (Watila & Balarabe, 2015). Stroke and language related variables, including initial aphasia severity, lesion size and location, and time post onset, have been identified as key predictors of recovery (Basso, 1992; Hope et al., 2021; Iorga et al., 2021; Lazar et al., 2010; Pedersen et al., 1995; Plowman et al., 2012; REhabilitation and recovery of peopLE with Aphasia after StrokE (RELEASE) Collaborators, 2021). In contrast, demographic factors, such as age, sex and handedness, are not consistently reported to impact upon recovery (Basso, 1992; Plowman et al., 2012; REhabilitation and recovery of peopLE with Aphasia after StrokE (RELEASE) Collaborators, 2021). An emerging body of neuroimaging research has further contributed to our understanding of treatment-induced recovery of aphasia (Copland, 2020; Hope et al., 2021; Iorga et al., 2021; Meinzer & Breitenstein, 2008). Separate studies conducted by Billot et al. (2022) and Hope et al. (2021) have demonstrated that stroke-related variables, such brain structural integrity and functional connectivity, are positively related to treatment-induced recovery. However, interestingly, both studies found that the best predictive models incorporated behavioural data, including baseline aphasia severity and dose of intervention, respectively. These findings provide support for the ongoing consideration of behavioural factors with regards to predictors of aphasia recovery.

A recent review by Varkanitsa and Kiran (2022) identifies several factors, including stroke and language related variables (e.g., brain and lesion characteristics, aphasia severity), demographic variables (e.g., age) and treatment-related factors (e.g., treatment type and intensity) that may influence treatment-induced recovery of aphasia. There is growing evidence that individuals’ non-linguistic cognitive profiles, may influence anomia recovery (Gilmore et al., 2019; Lambon Ralph et al., 2010; Peñaloza et al., 2022). Increasingly, researchers are acknowledging the influence of cognitive factors (e.g., attention, memory, executive function) and an individual’s (re)learning capacity in the process of rehabilitation (Dignam et al., 2017; Ferguson, 1999; Gilmore et al., 2019; Lambon Ralph et al., 2010; Peñaloza et al., 2022; Vallila-Rohter & Kiran, 2013). Recent research has further investigated individuals’ neuropsychological profiles and treatment-related factors (e.g., treatment type, components of therapy tasks) in order to predict response to intervention (Evans et al., 2021; Gravier et al., 2018; Kristinsson et al., 2021).

With respect to language factors, initial aphasia severity and lexical-semantic comprehension have been found to significantly predict an individual’s response to anomia therapy (Evans et al., 2021; Iorga et al., 2021; Lambon Ralph et al., 2010; Paolucci et al., 2005). Martin and colleagues (Martin et al., 2004; Renvall et al., 2005; Renvall et al., 2003) found that the integrity of an individual’s lexical-semantic comprehension was an important predictor of anomia therapy gains in response to a contextual repetition priming treatment. These findings suggest that discrete language components, as opposed to a gross measure of aphasia severity, may differentially influence treatment response and may provide a measure of an individual’s potential treatment responsiveness.

Consideration of the role of discrete language skills may provide a clinically feasible, personalized method of predicting language recovery in response to anomia therapy. Schliep et al. (2021) conducted a small (n = 7), exploratory study to investigate the relationship between naming stimulability and spontaneous recovery of word retrieval skills over a 12-month period, post-stroke. In contrast to expectations, this study found that naming stimulability, as measured by improvements in word retrieval following a semantic or phonological cue, was not significantly correlated with naming performance at 6 weeks, 3, 6 months or 12 months post stroke onset. Furthermore, attempts at naming, as measured by the number of production attempts for incorrect items divided by the total number of errors and future naming accuracy, were not significantly correlated with future naming accuracy. While self-reported data regarding the type and amount of aphasia rehabilitation participants received during this period was captured, measuring treatment-specific performance was not an explicit aim of the study and naming accuracy in response to intervention was not measured. As such, the generalisability of these findings to predictors of treatment response in anomia for individuals with aphasia is limited.

In addition to baseline demographic, stroke and language-related predictors, measuring early behavioural treatment response may provide a simple method for predicting overall treatment outcome. In a study of 10 participants with post-stroke aphasia, Simic et al. (2020) investigated whether two treatment-related variables, phonemic cue responsiveness and within therapy improvement (early / late), predicted naming accuracy for treated and untreated stimuli at post-therapy, 4 and 8 weeks follow-up. Participants received phonological component analysis (PCA) treatment 3 times per week for 5 weeks. Both phonemic cue responsiveness and (early) within therapy improvement emerged as significant predictors of naming accuracy for treated items immediately post-therapy, however, not at 4 weeks or 8 weeks follow-up. Performance on cue responsiveness and within therapy improvement was predictive of naming accuracy for untreated items at 4 weeks follow-up but not at post-therapy or 8 weeks follow-up. Interestingly, this study found that the greatest improvement in naming accuracy occurred during the early phase of therapy, within the first four rounds of PCA therapy. These findings have important clinical implications as they suggest that an individual’s performance after a brief period of anomia therapy may predict overall response to intervention and as such, may provide an accessible and useful tool for clinicians to predict individual response to anomia treatment.

The primary aim of the present study is to investigate whether early within-therapy probe naming performance, as measured by confrontation naming accuracy on a set of 30 treated picture stimuli after 3 h of impairment therapy, predicts treatment response for treated and untreated stimuli in a large sample of adults with post-stroke aphasia. In order to address this aim, we first established anomia treatment response at the individual participant and group level in response to the Aphasia LIFT programme. It is hypothesized that overall participants will demonstrate improved accuracy of word retrieval in response to the Aphasia LIFT programme (Dignam et al., 2015) and that there will be inter-individual variability in participants’ response to treatment (Menahemi-Falkov et al., 2021). Simic et al. (2020) found that the majority of improvement in naming accuracy occurred within the early phase of treatment. As such, we predict that early within-therapy probe naming performance, collected after 3 h of impairment therapy (treatment naming probe 1) will significantly predict anomia treatment outcomes at post-therapy and 1 month follow-up. In addition to considering the relationship between early within-therapy probe naming performance and anomia therapy outcomes, we also explored the influence of clinically accessible, participant factors including stroke and language related factors (i.e., time post onset, baseline lexical-semantic processing ability), and demographics (i.e., age) on therapy response for treated and untreated items.

Methods

This study was conducted as part of a broader research programme investigating the clinical efficacy of the intensive, comprehensive aphasia therapy programme, Aphasia LIFT. Demographic and clinical assessment results for these participants’ are published in Dignam et al. (2015) and Dignam et al. (2017).

Participants

Thirty-four participants (28 M, 6 F; Mean age 58.5y, SD 10.9) with chronic, post-stroke aphasia resulting from unilateral, left hemisphere stroke/s were recruited to the study. Thirty-two participants completed the therapy trial (Table 1). Two D-LIFT participants (P29, P31) withdrew from the study prior to the completion of therapy due to acute onset illness, and their data have been excluded from analyses. One D-LIFT participant (P18) did not complete the 1 month follow-up assessment due to a change in personal circumstances, giving a sample size of n = 31 at follow-up. Participants were allocated to an intensive (LIFT, n = 16) or distributed (D-LIFT, n = 16) treatment condition based on their geographic location, the availability of a position within the research programme and personal factors (e.g., participant availability, transport, accommodation). All participants were > 4 months post-stroke (mean 38.7 mo, SD 50.4), spoke fluent English prior to their stroke and were diagnosed with aphasia based on a score of < 62.8 on the Comprehensive Aphasia Test (CAT; Swinburn et al., 2004). One participant (P33) with a borderline CAT aphasia severity score of 63.0 was included in the study due to the presence of significant word-finding difficulties in conversation. Participants with comorbid neurological impairments, severe apraxia of speech or dysarthria were excluded from the study.

Table 1. Sample Characteristics at Baseline.

Variable
Sample size	32
Sex	5F, 27M
Mean age (SD), y	58.1 (10.7)
Handedness (EHI), n
Right	29
Left	3
Location of Stroke (left hemisphere), n	32
Time Post Onset (SD), mo	40.3 (9.1)

Note. N = sample size; EHI = Edinburgh Handedness Index.

Assessment

Participants completed a language assessment battery, administered by a qualified speech pathologist, prior to commencing therapy. Participants’ receptive and expressive language skills were evaluated using the Comprehensive Aphasia Test (CAT; Swinburn et al., 2004) (Supplementary Table 1). An estimate of participants’ baseline lexical-semantic processing was obtained by taking the sum of participants’ single word auditory and written comprehension scores from the CAT. In order to identify targets for anomia therapy, three baseline naming probes were administered. Naming probes for treated and untreated items were administered at baseline, within-therapy, immediately post-therapy and at 1 month follow-up.

Confrontation naming probes

Baseline naming probes

Three baseline naming probes were administered using 309 picture stimuli (nouns) obtained from the Bank of Standardised Stimuli (BOSS) (Brodeur et al., 2010). Picture stimuli were displayed on a computer screen in a confrontation naming task and accurate responses and self-corrections produced within 10 s were scored as correct. A total of 48 items that the participant was unable to name accurately (0/3 or 1/3 on baseline assessment) was selected and randomly allocated to treated (n = 24) or untreated (n = 24) items. In order to provide a level of success with therapy, 12 items that the participant was able to name accurately (2/3 or 3/3 on baseline assessment) were also selected and randomly allocated to treated (n = 6) or untreated items (n = 6). Independent-samples t tests confirmed that treated (n = 30) and untreated control (n = 30) sets were comparable with respect to baseline naming accuracy, SUBTITLE frequency (Balota et al., 2007), name agreement (Brodeur et al., 2010), and number of syllables (p > .05).

Within-therapy naming probes

Confrontation naming accuracy for treated and untreated items was probed after every 3 h of impairment therapy. Within-therapy naming probes were administered at the start of the therapy session and participant feedback was not provided. The presentation of treated and untreated items was pseudorandomised within the presentation. Consistent with the baseline naming probes, an accurate response or self-correction produced within 10 s was scored as correct. A total of four within-therapy naming probes were administered during the Aphasia LIFT programme. Raw confrontation naming accuracy for treated items (n = 30) collected after the first 3 h of impairment therapy was used in analyses as the measure of early within-therapy probe naming performance (i.e., treatment naming probe 1).

Post-therapy and follow-up naming probes

Confrontation naming accuracy for treated and untreated items was measured immediately post-therapy and at 1 month follow-up. Consistent with baseline and within-therapy naming probes, treated and untreated items were pseudorandomised within the presentation and an accurate response or self-correction produced within 10 s was scored as correct.

Therapy

Therapy was administered in accordance with the principles of Aphasia LIFT outlined in Rodriguez et al. (2013). Participants each received 48 h of aphasia therapy, which predominantly targeted word retrieval impairments. Therapy was comprised of 14 h of impairment therapy, 14 h of computer therapy, 14 h of functional therapy and 6 h of psycho-social group therapy. Impairment therapy incorporated training of 30 treated items using semantic feature analysis and phonological components analysis (Boyle, 2010; Boyle & Coelho, 1995; Leonard et al., 2008). Computer therapy reinforced training of these items using the computer software programme StepbyStep (Steps Consulting Limited., 2002). Further details regarding the therapy procedures are reported in Dignam et al. (2016) and Dignam et al. (2015).

A comprehensive Aphasia LIFT manual was developed to promote treatment fidelity. Therapy was provided by qualified speech pathologists who received training on the treatment approaches used in Aphasia LIFT. In some instances, computer therapy was facilitated by trained speech pathology students or a trained allied health assistant under the supervision of a qualified speech pathologist. Treating speech pathologists met weekly during the intervention to review treatment approaches, discuss participant progress and to resolve potential clinical issues. The fidelity of the intervention was not formally assessed in this study.

Data analysis

The primary aim of the study was to investigate the relationship between baseline language variables (early, within-therapy probe naming and baseline lexical-semantic processing ability) and anomia therapy gains. The study sought to explore the relative contributions of these baseline language variables to therapy gains for treated and untreated items. To maintain consistency with the analytical approach applied in Dignam et al. (2017) and Dignam et al. (2016), multiple regression analyses were used to explore the relative contributions for baseline language variables, namely early within-therapy probe naming and lexical-semantic processing ability, on anomia therapy response.

Treatment outcomes

In order to evaluate the influence of treatment-related variables on therapy response, efficacy of the intervention was first established. Both individual and group-level analyses were conducted to evaluate changes in treated and untreated items in response to anomia therapy. Therapy outcomes for treated and untreated items were analyzed at the individual level using the WEighted STatsistics (WEST) method outlined in Howard et al. (2014). This analysis considers individual variability during the baseline phase and compares participants’ pretherapy naming accuracy with naming accuracy at posttherapy and 1-month follow-up using a weighted one-sample t test (Howard et al., 2014). To prevent potential bias and confounds of regression to the mean (Howard et al., 2014), six stimuli that participants were able to name accurately at baseline (2/3 or 3/3 accurate) were included in the analyses. A single outcome score for performance at treatment naming probe 1 was calculated using the proportion of potential maximal gain (e.g., treatment naming probe 1 raw score – pre-therapy mean score)/(total number of items – pre-therapy mean score) (Lambon Ralph et al., 2010).

Group-level data were analysed using Linear Mixed Models (LMM) with a sample of n = 32 at post-therapy and n = 31 at 1 month follow-up. Prior to analyses, data for treated items were transformed using reflect and square root transformations (Tabachnick & Fidell, 2007).¹ Data approximated a normal distribution according to the Shapiro–Wilk test (p > .05) (Shapiro & Wilk, 1965). In order to evaluate the effects of intervention, separate models were fitted for treated and untreated items with time (pre-therapy, post-therapy, follow-up) and group (LIFT, D-LIFT) as a fixed effect and participants as a random effect. Cohen’s d effect sizes were calculated to determine the magnitude of treatment effects (Figure 1; Cohen, 1988). Data analyses were conducted using IBM SPSS Statistics (Version 28) (IBM Corp., Released 2021).

Figure 1. Formula used to calculate Cohen’s d effect sizes for group-level data.

Formula used to calculate Cohen’s d effect size. Cohen’s d equals the effect size divided by the pooled standard deviation.

Predictors of treatment response

In order to establish a single treatment outcome score for treated and untreated items, the proportion of potential maximal gain was calculated at post-therapy and 1 month follow-up. Proportion of treatment gain at post-therapy was transformed using a reflect and logarithmic transformation (Tabachnick & Fidell, 2007).² The proportion of potential maximal gain for treated items at post-therapy (transformed) and 1 month follow-up, approximated a normal distribution according to the Shapiro Wilk test (p > .05) (Shapiro & Wilk, 1965). Prior to determining the relative strength and contribution of independent variables to anomia therapy outcomes, we first established the presence of a relationship between independent variables and the dependent variable using Pearson correlation analyses. Consistent with the broader programme of research (Dignam et al., 2016; Dignam et al., 2017), multiple regression analyses were conducted to determine the relative contributions of lexical-semantic processing ability at baseline and early within-therapy probe naming performance to naming accuracy for treated and untreated items at post-therapy and 1 month follow-up. Independent variables that were significantly correlated with treatment outcomes at post-therapy or 1 month follow-up were entered into the multiple regression analyses (Murray, 2012). To account for potential differences between treatment conditions, Group (i.e., LIFT/D-LIFT) was also entered into the multiple regression analyses. Prior to finalizing the multiple regression models, assumptions of normality, linearity, multi-collinearity and homoscedasticity of residuals were tested and met.

Results

Treatment outcomes

Individual outcomes

Twenty-six out of 32 participants made statistically significant improvements in confrontation naming accuracy for treated items at post-therapy and therapy gains were maintained for 21 out of 31 participants at 1 month follow-up. Furthermore, a significant improvement in naming accuracy for untreated items was found for nine out of 32 participants at post-therapy and was maintained for six out of 31 participants at follow-up. There was individual participant variability in treatment response and participants’ proportion of potential maximal gain for treated items and untreated items at treatment naming probe 1, post-therapy and 1 month follow-up are reported in Table 2 (nb. Proportion of potential maximal gain for treated and untreated items have previously been reported in Dignam et al. (2017)).

Table 2. Individual therapy outcomes for treated and untreated items.

		Treated Items Proportional Gain			Untreated Items Proportional Gain
ID	Group	Probe 1	Post-therapy	Follow-up	Probe 1	Post-therapy	Follow-up
1	LIFT	.74*	.82*	.56*	.05	.24	.10
2	LIFT	.39*	.43*	.27	.08	−.02	.12
3	LIFT	.38*	.57*	.45*	−.08	−.04	.21
4	LIFT	.25	.25	.18	.07	.07	.04
5	LIFT	.24	.33	.06	.15	.11	−.10
6	LIFT	.74*	.84*	.32	.17	.28	.38
7	LIFT	.73*	.77*	.61*	.47*	.26	.22
8	LIFT	.67*	.96*	.63*	.29	.25	.21
9	LIFT	.91*	.87*	.78*	.54*	.45*	.45*
10	LIFT	.30*	.34*	.26*	.05	.10	.10
11	LIFT	.35*	.51*	.23	.07	.12	−.01
12	LIFT	.79*	.91*	.66*	.26	.43*	.30
13	LIFT	−.09	.30	.39	.01	−.07	.10
14	LIFT	.87*	.96*	.91*	.28	.42	.33
15	LIFT	.15	.83*	.58*	.07	.11	.24
16	LIFT	.10	.10	.03	.06	−.01	−.01
17	D-LIFT	.21	.21	.16	.03	.15	.07
18	D-LIFT	.95*	.91*	N/A	.45	.55	N/A
19	D-LIFT	.12	.40*	.08	.01	−.11	−.11
20	D-LIFT	.42*	.58*	.42*	−.03	.15	.24
21	D-LIFT	.75*	.95*	.80*	.12	.26	.22
22	D-LIFT	.79*	1.00*	.95*	.62*	.67*	.43*
23	D-LIFT	−.12	.06	.06	−.17	.02	.02
24	D-LIFT	.38*	.43*	.57*	.24	.29	.14
25	D-LIFT	.50*	.82*	.64*	.44	.54*	.44
26	D-LIFT	.67*	.71*	.43*	.08	.26	.22
27	D-LIFT	.22	.77*	.63*	−.05	.05	−.05
28	D-LIFT	.56	.81*	.76*	.30	.60*	.40*
30	D-LIFT	.85*	.95*	.95*	.21	.70*	.85*
32	D-LIFT	.77*	.82*	.68*	.23	.44*	.23
33	D-LIFT	.75*	.85*	.69*	.34*	.37*	.47*
34	D-LIFT	.68*	.91*	.91*	.38*	.43*	.38*

Note. ID = Participant identification; LIFT = Intensive treatment condition; D-LIFT = Distributed treatment condition; * WEST-ROC Analysis p < .05.

Group-level outcomes

At the group level, LMM revealed a significant increase in confrontation naming accuracy for treated items at post-therapy, F(1,30) = 122.76, p < .001, d = 2.7, and 1 month follow-up, F(1,29) = 79.55, p < .001, d = 2.1, compared with pre-therapy. Furthermore, there was a significant increase in confrontation naming accuracy for untreated items at post-therapy, F(1,30) = 47.09, p < .001, d = 1.2, and follow-up F(1,29) = 36.22, p < .001, d = 1.1, compared with pre-therapy. There was no significant time by group (LIFT/D-LIFT) interaction for treated or untreated items at post-therapy or 1 month follow-up (p > .05)

Early within-therapy probe naming performance and predictors of treatment response

Early within-therapy probe naming performance

Twenty-two out of 32 participants demonstrated statistically significant improvements in naming accuracy for treated items at treatment naming probe 1, according to individual WEST analyses. Proportion of potential maximal gain for treatment naming probe 1 is presented in Table 2.

Pearson correlations

Pearson correlation analyses between baseline demographics, lexical-semantic processing ability, treatment naming probe 1 and therapy outcomes for treated and untreated items at post-therapy and 1 month follow-up are reported in Table 3. There was no significant correlation between baseline demographic variables, namely age and time post onset, and therapy gains for treated and untreated items at post-therapy or 1 month follow-up. As such, these baseline demographic variables were excluded from further analyses.

Table 3. Pearson correlations for language variables and therapy gains for treated and untreated items.

	Treated Items		Untreated Items
	Post Therapy	Follow-up	Post Therapy	Follow-up
Lexical-semantics	−.666**	.665**	.489**	.589**
Treatment naming probe 1	−.850**	.792**	.827**	.722**
Age	.168	−.172	−.054	−.015
Time Post Onset	−.059	−.051	−.052	−.145

Note. *p < .05; **p < .01.

Multiple regression analyses

Treated items

Three variables were entered into the multiple regression model to establish the relationship between baseline lexical-semantic processing ability, early within-therapy probe naming performance and therapy gains for treated items at post-therapy (group, lexical-semantics, treatment naming probe 1). The multiple regression model was statistically significant and accounted for 76.5% of the variance in anomia treatment outcomes at post-therapy, R² = .765, adjusted R² = .740, F(3, 28) = 30.37, p < .001 (Table 4). The beta weights indicate that early within-therapy probe naming performance, β = −.686, p < .001, and lexical-semantic processing, β = −.260, p = .034, significantly contributed to gains in naming accuracy for treated items at post-therapy. Furthermore, squared semi-partial correlations indicate that 28.3% of the variance was uniquely accounted for by early within-therapy probe naming performance, whereas lexical-semantic processing contributed 4.2%.

Table 4. Multiple regression model with proportion of potential maximal therapy gain for treated items at post therapy and 1 month follow-up as the dependent variable.

		Regression Coefficient (B)	Standard Error (B)	95% Confidence Interval (B)		Standardized Coefficient (β)	Squared semipartial correlations (sr^2)	t	p value
				Lower	Upper
Post Therapy	Group	−.030	.055	−.142	.083	−.051	.002	−.540	.594
	Probe 1*	−.027	.005	−.036	−.017	−.686	.283	−5.80	<.001
	Lexical-semantics	−.007	.003	−.013	−.001	−.260	.042	−2.23	.034
1month Follow-up	Group	.121	.057	.005	.238	.218	.045	2.13	.042
	Probe 1*	.021	.005	.011	.031	.551	.189	4.39	<001
	Lexical-semantics	.009	.003	.003	.015	.361	.082	2.89	.007

*Treatment naming probe 1.

Three variables were entered into the multiple regression model to establish the relationship between baseline lexical-semantic processing ability, early within-therapy probe naming performance and therapy gains for treated items at 1 month follow-up (group, lexical-semantics, treatment naming probe 1). The multiple regression model was statistically significant and accounted for 73.5% of the variance in anomia treatment outcomes at 1 month follow-up, R² = .735, adjusted R² = .705, F(3, 27) = 24.94, p < .001 (Table 4). The beta weights indicate that early within-therapy probe naming performance, β = .551, p < .001, and lexical-semantic processing, β = .361, p = .007, significantly contributed to gains in naming accuracy for treated items at 1 month follow-up. Furthermore, squared semi-partial correlations indicate that 18.9% of the variance was uniquely accounted for by early within-therapy probe naming performance, whereas lexical-semantic processing contributed 8.2%.

Untreated items

Three variables were entered into the multiple regression model to establish the relationship between baseline lexical-semantic processing ability, early within-therapy probe naming performance and naming accuracy for untreated items at post-therapy (group, lexical-semantics, treatment naming probe 1). The multiple regression model was statistically significant and accounted for 73.9% of the variance in anomia treatment outcomes at post-therapy, R² = .739, adjusted R² = .711, F(3, 28) = 26.5, p < .001 (Table 5). The beta weights indicate that early within-therapy probe naming performance, β = .749, p < .001, significantly contributed to naming accuracy for untreated items at post-therapy and accounted for 33.6% of the unique variance in naming performance.

Table 5. Multiple regression model with proportion of potential maximal therapy gain for untreated items at post therapy and 1 month follow-up as the dependent variable.

		Regression Coefficient (B)	Standard Error (B)	95% Confidence Interval (B)		Standardized Coefficient (β)	Squared semipartial correlations (sr^2)	t	p value
				Lower	Upper
Post Therapy	Group	.108	.044	.018	.198	.245	.056	2.45	.021
	Probe 1*	.022	.004	.015	.030	.749	.336	6.01	<.001
	Lexical-semantics	.001	.002	−.004	.006	.055	.002	.450	.656
1month Follow-up	Group	.076	.051	−.028	.180	.188	.033	1.46	.146
	Probe 1*	.014	.004	.006	.023	.521	.169	3.37	.002
	Lexical-semantics	.005	.003	.000	.011	.299	.057	1.95	.062

*Treatment naming probe 1.

Three variables were entered into the multiple regression model to establish the relationship between baseline lexical-semantic processing ability, early within-therapy probe performance and naming accuracy for untreated items at 1 month follow-up (group, lexical-semantics, treatment naming probe 1). The multiple regression model was statistically significant and accounted for 59.7% of the variance in anomia treatment outcomes at 1 month follow-up, R² = .597, adjusted R² = .553, F(3, 27) = 13.36, p < .001 (Table 5). The beta weights indicate that early within-therapy probe naming performance significantly contributed to therapy outcome at 1 month follow-up, β = .521, p = .002, and uniquely accounted for 16.9% of the variance in performance.

Discussion

In view of the heterogenous nature of aphasia and individuals’ variable response to anomia therapy, a greater understanding of the language and participant-related factors that may influence individual treatment response is necessary to inform clinical practice. This study evaluated anomia therapy outcomes in response to the intensive, comprehensive aphasia programme, Aphasia LIFT. The recent findings of Menahemi-Falkov et al. (2021) indicate that group level analyses may mask the variability of individuals’ response to aphasia therapy and as such, we investigated treatment response at the group and individual level. At the group level, we found significant improvements in naming accuracy for treated and untreated items at post-therapy(Cohen, 1988) and these effects were maintained at 1 month follow-up. Gains for treated and untreated items at post-therapy and 1 month follow-up were associated with a large effect size (d > .80). Performance on an early within-therapy confrontation naming probe, administered after 3 h of impairment therapy, was predictive of gains in confrontation naming for treated and untreated items at post-therapy and 1 month follow-up. In addition, baseline lexical-semantic processing ability also emerged as a significant predictor of anomia therapy response for treated items at post-therapy and 1 month follow-up.

While there was individual variability in treatment outcomes, we found that the majority of participants (26 out of 32 participants) achieved statistically significant gains in naming accuracy for treated items post-therapy and this was maintained for approximately two thirds of participants (21 out of 31 participants) at 1 month follow-up. Furthermore, there was evidence supporting the generalization of treatment effects to untreated items for a small number of participants at post-therapy (nine out of 32 participants) and this was maintained for six participants at follow-up. The findings of the present study, focused on treated and untreated items, are consistent with the results of Dignam et al. (2015) and Rodriguez et al. (2013), which found a positive and enduring treatment effect for Aphasia LIFT on the Boston Naming Test (Kaplan et al., 2001). Thus our current findings provide further support for the efficacy of Aphasia LIFT in the remediation of word retrieval deficits.

Given the variability in individuals’ response to aphasia therapy, it is clinically important to be able to identify who will respond to intervention and to predict the potential magnitude of their treatment response. This study found that early within therapy probe naming performance, in conjunction with a baseline measure of lexical-semantic processing ability, significantly predicted improvements in naming accuracy for treated and untreated items in response to the Aphasia LIFT programme. Treatment naming probe 1 emerged as a significant, independent predictor of naming accuracy for treated items at post-therapy, accounting for 28.3% of the unique variance in treatment response. At the individual level, of the 26 participants who demonstrated significant gains in naming accuracy for treated items at post-therapy, 22 of these participants demonstrated significant naming improvements at treatment naming probe 1, after just 3 h of impairment therapy. These findings are consistent with Simic et al. (2020), in that we found statistically significant gains in naming accuracy occurred early within the treatment period for the majority of treatment responders and that early gains in naming accuracy for treated items were predictive of improved naming accuracy for treated items at post-therapy. In contrast to Simic et al. (2020), we found that naming accuracy at treatment naming probe 1 was also predictive of performance for treated items at 1 month follow-up, accounting for 18.9% of the variance in treatment response. Of the 22 participants who achieved early treatment response at treatment naming probe 1, 18 out of 21 participants maintained these gains for treated items at 1 month follow-up (>85%). Together, these findings build upon the work of Simic et al. (2020) and provide further support for the value of early treatment response as a tool for prognostication in anomia rehabilitation.

Whilst only a small number of participants demonstrated generalization of treatment effects to untreated items at post therapy and 1 month follow-up, naming performance for treated items at treatment naming probe 1 also emerged as a significant predictor of generalization, accounting for 33% and 16.9% of the variance at post-therapy and 1 month follow-up, respectively. At the individual level, of the participants who demonstrated significant gains in naming accuracy for untreated items, 8 out of 9 participants at post therapy and 5 out of 6 participants at 1 month follow-up demonstrated significant improvements in naming accuracy at treatment naming probe 1. Thus, consistent with Simic et al. (2020), early within therapy improvement may be indicative of treatment responsiveness for both treated and untreated items. Conversely, it is also important to acknowledge that 17 out of 22 participants who demonstrated significant improvements for treated items at treatment naming probe 1 did not go on to make gains in untreated items.

It is worth noting that baseline lexical-semantic processing ability emerged as a significant predictor of therapy response for treated items, independent of early within therapy probe naming performance. Previous research has identified discrete language skills, including auditory and reading comprehension and semantic and phonological processing skills, as significant predictors of anomia treatment response (Gilmore et al., 2019; Iorga et al., 2021; Kristinsson et al., 2021). Furthermore, this finding is consistent with previous reports in the literature regarding the role of lexical-semantic processing in word (re)learning (Martin et al., 2004; Renvall et al., 2005; Renvall et al., 2003). While both baseline lexical-semantic ability and probe naming performance emerged as independent predictors for treated items, probe naming performance accounted for greater variance in treatment response. Within therapy probe naming performance aligns closely with the focus of the treatment, with regards to task specificity, and therefore may be a more sensitive predictor of treatment response. The relative contributions of early within therapy probe naming performance and lexical-semantic processing to anomia therapy gains for treated items is also of interest. While early within therapy probe naming accounts for 28.3% of the variance in therapy gains for treated items at post-therapy, this is reduced to 18.9% at 1 month follow-up. Conversely, lexical-semantic processing accounts for 4.2% of the variance in gains for treated item at post-therapy, but increases to 8.2% at 1 month follow-up. Peñaloza et al. (2022) provides a review of theories and mechanisms of learning in people with aphasia. It is acknowledged that language learning occurs in phases (e.g., initial encoding, consolidation) and that these phases may be mediated by different cognitive and language processing mechanisms (Peñaloza et al., 2022). As such, it is suggested that probe naming performance potentially captures elements of an individual’s learning potential, including short-term memory processes, and engagement in rehabilitation that is not evident in baseline performance (i.e., lexical-semantic processing) alone, and that these processes may be relevant for the initial acquisition of treatment gains, thus accounting for the greater variance in treatment outcome at post-therapy. In contrast, language processing mechanisms, as measured by lexical-semantic processing, may be particularly relevant for moderating the long-term maintenance of treatment gains, as indicated by the increased relative contribution of lexical-semantic processing on treatment outcomes at 1 month follow-up.

Previous research has considered potential markers of responsivity to aphasia therapy, including cognitive and learning capacity and it is of interest to consider which of these factors may best predict individuals’ recovery. As part of the broader programme of research, Dignam et al. (2016) found that novel word learning and baseline lexical-semantic processing ability significantly predicted anomia treatment outcomes at post-therapy, accounting for 49.1% of the variance in treatment outcomes. In contrast, the present study found that baseline lexical-semantic processing and early within-therapy probe naming performance, collected after 3 h of intervention, accounted for 76.5% of variance in treatment outcomes at post-therapy, above that of a measure of novel word learning capacity at baseline alone. Interestingly, the model incorporating cognitive factors, verbal short-term memory and learning ability, and baseline lexical semantic processing, reported in Dignam et al. (2017), accounted for 72.3% of variance in treatment outcomes at post-therapy, similar to the present findings. As discussed above, it is possible that performance on the early within therapy naming probe provides an informal measure of participants’ cognitive and learning ability, which is important for the initial acquisition of treatment gains. The treatment-induced recovery of anomia is acknowledged to be multifactorial and many behavioural characteristics, including cognitive and language function as well as personal characteristics such as motivation, self-awareness and resilience, may influence treatment response. Further research is required to determine which of these characteristics are critical and how this information may be used clinically to determine potential treatment responsiveness.

Interestingly, while lexical-semantic processing emerged as a significant, independent predictor of therapy gains for treated items at post-therapy and 1 month follow-up, it did not emerge as a significant predictor of gains for untreated items. Consistent with accounts from Howard et al. (2014) and Nickels (2002a), our previous work suggested that repeated exposure to naming probes may have resulted in improved naming performance for untreated items and that this effect was particularly evident in individuals with superior short-term memory and learning processes (Dignam et al., 2017). Whilst the mechanisms of generalization of anomia therapy to untreated items remains unknown, the present results provide support for the differential influence of language and cognitive processes to the acquisition and maintenance of treated versus untreated items.

There has previously been debate in the literature regarding the influence of demographic variables, such as age, on rehabilitation outcomes for aphasia (Basso, 1992; Plowman et al., 2012; REhabilitation and recovery of peopLE with Aphasia after StrokE (RELEASE) Collaborators, 2021; Watila & Balarabe, 2015). The neuroscience literature suggests that training induced plasticity occurs more readily in younger brains (i.e., Age Matters) (Kleim & Jones, 2008) and this finding has been reflected in the recent meta-analyses conducted by REhabilitation and recovery of peopLE with Aphasia after StrokE (RELEASE) Collaborators (2021). REhabilitation and recovery of peopLE with Aphasia after StrokE (RELEASE) Collaborators (2021) found that younger stroke survivors (< 55 years) exhibited greater gains across all language domains; however, clinical gains were still evident in older stroke survivors (> 75 years). We did not find that age was significantly correlated with treatment outcomes in the present study, although it should be noted that the mean participant age in our sample was 58.5 years and our sample did not include adults over 77 years.

We also found that therapy outcomes in the present study were not significantly correlated with time post onset. While there may be some clinical benefits of commencing aphasia therapy earlier in the subacute phase of stroke recovery (REhabilitation and recovery of peopLE with Aphasia after StrokE (RELEASE) Collaborators, 2021), our findings demonstrate that significant gains can still be made in the chronic phase. Encouragingly, participants as long as 18 years post-stroke demonstrated a positive response to Aphasia LIFT. This finding is further supported by the review conducted by Allen et al. (2012) which asserts that aphasia therapy initiated greater than 6 months post stroke onset can be effective.

Whilst not a primary aim of this study, it is important to consider the potential influence of treatment intensity on anomia therapy outcomes. Aphasia LIFT was provided in either an intensive versus distributed treatment schedule and treatment group was included as a fixed factor in the linear mixed model analyses. Importantly, we found no time x group interaction for treated or untreated items at post-therapy or 1 month follow-up, providing support for the efficacy of Aphasia LIFT when delivered in both an intensive and distributed treatment schedule.

A greater understanding of the influence of language and participant-related factors on response to anomia therapy has important implications for clinical practice; however, it also requires ethical consideration. Early within-therapy probe naming, is an accessible, cost-effective tool that may support prognostication, clinical decision making, and service provision in aphasia rehabilitation. Information obtained from early within-therapy naming probes may also supplement speech pathologists’ existing clinical impression of a patient, obtained from formal language assessment. For example, within-therapy probe naming performance may provide an indirect measure of the individuals’ cognitive and personal capacity to engage in rehabilitation for word retrieval deficits. However, the information obtained from within-therapy, probe naming assessments alone is insufficient to make decisions regarding an individual’s suitability for rehabilitation. Clinical aphasia management is multifactorial, and speech pathologists must consider all elements of a patient’s clinical presentation (i.e., diagnosis, demographics, personal factors) when developing a management plan. Early within-therapy probe naming assessment is not intended as a tool to discriminate who does and does not receive therapy nor to determine when anomia therapy should be discontinued. Instead, the early identification of potential treatment response may support clinicians to counsel individuals with respect to their prognosis for a specific treatment. It may also aid in the selection of indicated treatment techniques, both by modifying existing treatment approaches to facilitate success or alternatively by introducing a new treatment approach. Speech pathologists are routinely required to consider these complex clinical and ethical considerations and adapt their clinical practice in the development of a wholistic management plan, accordingly.

While we observed significant improvements in naming accuracy after 3 h of impairment therapy, it is not suggested that only 3 h of therapy is required to maintain treatment gains long-term. Neuroscience research indicates that continued practice and over-learning of a skill, beyond the initial skill acquisition, is required to achieve neural reorganization and ultimately maintain performance of the skill over time (Kleim & Jones, 2008). Research conducted by Menahemi-Falkov et al. (2021) has highlighted the variability of individual treatment response to aphasia rehabilitation and has also highlighted the need for further investigation into factors that promote the long-term maintenance of treatment gains. Factors such as overall treatment dose, distribution of practice and task difficulty may influence the long-term maintenance of therapy gains and, therefore, are an important consideration for future research (REhabilitation and recovery of peopLE with Aphasia after StrokE (RELEASE) Collaborators, 2022). Additionally, the learning trajectory observed in the present study is consistent with learning algorithms reported in cognitive skill acquisition (e.g., Rescorla-Wagner theory, delta rule, power law of practice, Anderson, 1995); however, a greater understanding of the optimal dose and scheduling of intervention is required in order to maximize communication gains long-term and should be addressed in future research (Harvey et al., 2020; Menahemi-Falkov et al., 2021; REhabilitation and recovery of peopLE with Aphasia after StrokE (RELEASE) Collaborators, 2022).

The intervention provided in this study, Aphasia LIFT, is a complex intervention comprised of impairment, computer, functional and group-based therapy. Whilst impairment and computer therapy aimed to directly remediate confrontation naming of treated items, it is possible that therapeutic response may have been influenced by additional therapy components. Furthermore, consideration of the semantic and phonological relationship between treated and untreated items, and its’ potential confound on treatment response was beyond the scope of this study. As such, further studies investigating early treatment response and controlling for treatment type and psycholinguistic features of the stimuli are required.

Previous research has considered the role of discrete language components, including phonological processing and naming stimulability on spontaneous recovery and anomia treatment response (El Hachioui et al., 2013; Schliep et al., 2021). The present study considered the influence of lexical-semantic processing on anomia therapy outcomes; however, consideration of further discrete language skills was beyond the scope of the study. It would be of interest for future studies to investigate and compare the influence of a range of discrete, baseline language variables, including lexical-semantic and phonological processing as well as naming stimulability, on anomia treatment response.

Summary and conclusions

This study investigated the relationship between early within-therapy probe naming performance, collected after 3 h of impairment therapy, and anomia therapy outcomes at post-therapy and 1 month follow-up in adults with chronic, post-stroke aphasia. We also considered participants’ demographic and language profiles to identify further behavioural predictors of anomia treatment response. We found that early within-therapy probe naming performance significantly predicted anomia therapy outcomes for treated and untreated items at post-therapy and 1 month follow-up. In addition to within-therapy probe naming performance, baseline lexical-semantic processing ability but not age nor time post onset, emerged as a significant predictor of treatment response. This study has advanced our understanding of the relationship between early therapy response, participant characteristics and anomia therapy outcomes and has important clinical implications. Developing a clinically feasible, accessible method for identifying potential response to anomia therapy may assist clinical decision making and service provision and ultimately maximize therapy outcomes for adults with aphasia.

Acknowledgements

This work was supported by the National Health and Medical Research Council (NHMRC) Centre of Clinical Research Excellence in Aphasia Rehabilitation under Grant # 569935, a Royal Brisbane & Women’s Hospital Foundation grant and a Speech Pathology Australia post-graduate research grant. DC was supported by an Australian Research Council (ARC) Future Fellowship.

The Communication Research Registry is acknowledged as a source of participant recruitment. We would like to acknowledge the contribution of Eril McKinnon and Dr Asad Khan in the collection and statistical analysis of the data, respectively. We would also like to acknowledge the support provided by the research sites including; Prince of Wales Hospital (Randwick, NSW), the Royal Brisbane and Women’s Hospital (Herston, QLD), Royal Rehabilitation (Ryde, NSW) and St George Hospital (Kogarah, NSW). Finally, we would like to acknowledge the people with aphasia and their family members for participating in the programme.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data Availability

As requested by the reviewers, the dataset for this study has now been published with UQ eSpace (https://espace.library.uq.edu.au). Please advise on the most appropriate way to cite/refer to this dataset. The reference for this dataset is: Dignam, J., Rodriguez, A.D., O'Brien, K., Burfein, P., Copland, D. (2023) Early within therapy naming probes as a clinically-feasible predictor of anomia treatment response. The University of Queensland. Data Collection. https://doi.org/10.48610/1ae3b01

Additional information

Funding

This work was supported by National Health and Medical Research Council: [Grant Number Centre of Clinical Research Excellence in Aphasia].

Notes

1 Reflect and square root transformation calculated using the formula: f (x) =  √(31-x).

2 Reflect and logarithmic transformation calculated using the formula: f(x) = lg10(1.1-x)