Rehabilitation of post-stroke aphasia by a single protocol targeting phonological, lexical, and semantic deficits with speech output tasks: a randomized controlled trial

Abstract

BACKGROUND

The defective spoken output of persons with aphasia has anomia as a main clinical manifestation. Improving anomia is therefore a main goal of any language treatment.

AIM

This study assessed the effectiveness of a novel, 2-week, rehabilitation protocol (PHOLEXSEM), focused on PHonological, SEmantic, and LExical deficits, aiming at improving lexical retrieval, and, generally, spoken output.

DESIGN

A prospective, randomized controlled trial.

SETTING

In-patient and out-patient population of the Neurorehabilitation Unit of the Istituto Auxologico Italiano IRCCS, Milan, Italy.

POPULATION

The sample comprised 44 adults with aphasia due to left brain damage; 22 of them were assigned to the experimental (PHOLEXSEM) group, whereas 22 were assigned to the control group that received the Promoting Aphasics Communicative Effectiveness (PACE) protocol.

METHODS

All participants were treated 30-min daily for two weeks. The PHOLEXSEM training included 3 sets of exercises: 1) non-word, word, and phrase repetition; 2) semantic feature analysis by naming; 3) phonemic, semantic, and verb recall. Treatment effects were evaluated with tasks and items different from those used for training, to assess generalization effects.

RESULTS

After the PHOLEXSEM treatment, repetition, naming, lexical retrieval and sentence comprehension improved more than in the control – PACE – group, with gains generalizing to non-trained items. These improvements were independent of aphasia chronicity and only marginally influenced by demographic factors.

CONCLUSIONS

The 2-week PHOLEXSEM training, by targeting spoken output, ameliorates different aspects of aphasia, ranging from speech production (i.e., phonology and lexical retrieval) to comprehension.

CLINICAL REHABILITATION IMPACT

The PHOLEXSEM training is a useful and easy-to-administer intervention to improve post-stroke language deficits in adults of different ages, levels of education, duration, type, and severity of aphasia.

Aphasia is “...the more or less complete loss of the ability to use language”^{1, 2} resulting from acquired brain damage, typically of the left hemisphere. Aphasia comprises different deficits, which may occur both in association or independently of each another.³ Its severity varies from impairment of all functions (i.e., global aphasia) to deficits of specific components (phonology, lexicon, semantics, syntax, but with variable patterns of association among them) affecting language production, comprehension, or both.^{4, 5} Pragmatic skills may also be affected, further impairing efficient communication.¹

Aphasia rehabilitation has been inspired by cognitive models of language, according to a restorative approach focused on the treatment of the damaged mechanisms.^{1, 6} Impairment-based, restorative therapies aim at improving language deficits based on a functional diagnosis of the damaged mechanism. Conversely, communication-based (or consequence-based) therapies aim at enhancing communication, including non-verbal communication (e.g., by gestures and drawing), rather than specifically targeting the defective linguistic skills. These approaches include the Promoting Aphasics Communicative Effectiveness (PACE⁷) and the Supported Conversation Therapy.⁸ More recently, a mixed model has been proposed whereby restorative activities can be combined in a more ecological approach depending on the time elapsed since the injury.⁹ Such integrated approaches, which combine both restorative and communication-based paradigms, seem particularly useful since speech and language impairments have a strong impact on functional communication in everyday life. Although research in this field is evolving,¹⁰ the most updated Cochrane review on post-stroke aphasia rehabilitation¹¹ concludes that, whereas there is general evidence for the effectiveness of Speech and Language Therapy (SLT), this is still insufficient “to draw any conclusion regarding the effectiveness of one specific SLT approach over another”.

Anomia, namely the difficulty in retrieving words, is a main clinical symptom of all forms of aphasia^{12, 13} and its severity correlates with a low quality of life;¹⁴ hence, anomia represents a main target of SLT. Within the theoretical frameworks outlined above, exercises for the rehabilitation of anomia can be classified into two types: 1) substitutive treatments that promote the use of spared linguistic and nonlinguistic functions, or to circumvent the impairment through residual abilities (e.g., gestures or writing¹⁵); 2) restitutive, impairment-based, treatments that aim to reactivate lexical-semantic and phonological representations to improve word retrieval. Following the latter approach, naming exercises have long been the main activity used by therapists to improve lexical deficits; by using different cues, naming training have been shown to be effective in increasing the patients’ naming ability, particularly for treated items.^16-18 Most rehabilitation studies on anomia have focused primarily on noun retrieval, whereas verb retrieval has been less investigated so far.¹⁵

Among restorative treatments, the Semantic Feature Analysis (SFA) is a method for activating semantic networks. According to the review by Maddy et al.,¹⁹ the SFA is effective in improving anomia in both fluent and non-fluent aphasic individuals. However, there is still a lack of evidence for a generalization to other linguistic abilities of the improvements brought about by naming exercises.³

Many theories of language production agree that naming ability relies on activation spreading from the conceptual level to the lemma (which contains syntactic information such as the gender categories of a noun), and subsequently to the word form. The latter is hypothesized to include two separate mechanisms retrieving the phonological segments and the word structure (i.e., the syntactic rules). Discrete models²⁰ assume a unidirectional process in which a given representation (e.g., lemma access) must be selected before the activation of the subsequent representation (e.g., phonological access). Interactive models, instead, assume that functional activation spreads from semantics to word and phoneme units, and vice-versa.²¹ With regard to aphasia rehabilitation, it has been shown that semantic and phonological treatments can facilitate object and action naming in the same patient, although in different patients one treatment may be more effective than another.^{22, 23}

Based on these premises, the present study aimed at assessing the efficacy of a novel treatment for the rehabilitation of PHonological, LEXical, and SEMantic deficits (PHOLEXSEM) in persons with aphasia following an impairment-based perspective.³ With reference to the Interactive Two-step Model of lexical access^{24, 25} and the emphasis on the joint training of both phonological and lexical-semantic abilities for the rehabilitation of aphasia,³ the PHOLEXSEM treatment targets different aspects of aphasic speech output, namely: naming, repetition, and fluency. More specifically, the treatment involves: 1) repetition exercises with lexical and non-lexical stimuli of different lengths, known to improve both phonological output and lexical retrieval;^{13, 26, 27} 2) SFA to improve lexical retrieval through the analysis of the semantic features of the stimulus, with generalization effects to untreated items;²⁸ 3) activation of lexical retrieval by providing patients with phonological, semantic, and grammatical cues. It should be noted that the PHOLEXSEM treatment does not take into account the specific linguistic deficits of a person. The only requirement is the sparing of repetition skills. In fact, the underlying assumption of this approach is that an intervention targeting a core symptom of aphasia (i.e., anomia) may be useful before defying the prevailing pattern of the linguistic impairment (i.e., phonological, lexical-semantic, syntactic or pragmatic).

The clinical efficacy of a 2-week PHOLEXSEM treatment was evaluated in a sample of aphasic participants following a left-sided brain damage, in chronic or subacute stages of illness, by assessing post-training improvements with standardized neuropsychological tests for receptive and expressive language disorders. The improvements induced by the PHOLEXSEM treatment were compared to those achieved by a control group of persons with aphasia who received a 2-week PACE treatment. The PACE, although not targeting specific linguistic features as the PHOLEXSEM, was chosen because of its flexibility (indeed it can be used regardless of the type and severity of the aphasia), and its feasibility in the conventional rehabilitation setting, i.e., individual sessions in a limited amount of time. During the PACE, stimuli of increasing difficulty, selected between living and non-living, and sorted by frequency of use were presented. We discarded other therapies, such as the constraint-induced language therapy (CILT), requiring an intervention at a group level and a more intensive training as compared to our 2-week experimental protocol.²⁹

Materials and methods

Sample size estimation

The sample size was estimated by means of a power analysis (G*Power 3.1.9.6³⁰) for the interaction of between- and within-subject factors in a repeated-measures Analysis of Variance (ANOVA). We specified a medium effect size Cohen’s f=0.15 (a more conservative estimate than other works in the field, e.g., Stahl et al.³¹), α=0.05; 1-β=0.95; number of repeated measures: 4 assessment timepoints (see below) and 2 groups of persons with aphasia (i.e., the experimental and the control group); correlation between repeated measure=0.8, resulting in N.=40. Assuming a possible dropout rate of 10%, the final sample included 44 persons with aphasia.

Participants

Forty-four left-brain-damaged, right-handed, persons with aphasia were recruited from the in-patient and out-patient population of the Neurorehabilitation Unit of the IRCCS Istituto Auxologico Italiano, Milan, Italy (Table I).

Table I. —Demographic, clinical, and lesion data.

Variable	PHOLEXSEM	PACE
N.	22	21
Age (years)	67.5±9.49 (51-83)	65.1±12.8 (34-80)
Education (years)	13.6±3.98 (5-18)	12.9±4.37 (5-18)
Sex (F/M)	11/11	12/9
Stroke Type (I/H/I+H)	14/7/1	13/7/1
Disease duration (days)	114±104 (16-455)	60.1±96 (24-381)
Lesion volume (cm3)	32.9±20.1 (10.9-87)	63.3±55.3 (2.2-183)
FIM Score	49.1±21 (18-81)	88.1±25.4 (37-121)
Aphasia Type (F/NF)	5/17	6/15

Mean values, standard deviation and range are reported for age, education, disease duration, lesion volume, and Functional Independence Measure (FIM) Score. F: female; M: male; I: ischemic stroke; H: hemorrhagic stroke; F: fluent; NF: non-fluent.

The inclusion criteria were: being affected by a left-brain damage resulting in aphasia clinically confirmed by a neuropsychologist or a speech therapist; subacute or chronic stage of illness. Exclusion criteria were: having global aphasia and/or being enrolled in another treatment for aphasia.

All patients in a subacute stage of illness had not previously received aphasia rehabilitation, whereas those in a chronic stage of illness had received some treatment of aphasia in other hospitals that, to our knowledge, was not based on any specific protocol.

All participants agreed to be involved in the rehabilitation program and were informed that the set of exercises aimed at improving speech output or pragmatic efficacy. All of them provided their informed consent. The study was approved by the hospital Ethics Committee (approval code: 2015_03_31_03) and conducted in compliance with the ethical standards of the Declaration of Helsinki.

Neuropsychological assessment and study endpoints

The assessment was carried out with the following schedule: 1) 7 days before the beginning of the treatment (“Pre 1”); 2) on the first day, before the beginning of the rehabilitation session (“Pre 2”); 3) at the end of the first week of treatment (“Post 1”); 4) at the end of the second week of treatment (“Post 2”). The inclusion of a double baseline assessment (i.e., Pre 1 and Pre 2) was intended to control for possible practice effects and spontaneous recovery.

The pre- and post-treatment assessments were performed by E.B. and the treatment by C.C.

At each timepoint, the aphasia assessment comprised the following standardized tests, administered in random order. The Token Test was used to assess auditory language comprehension.^32-34 The language deficits were evaluated with the Italian test battery “Esame Neuropsicologico per l’Afasia” (ENPA),³⁵ which is devised to assess different linguistic deficits. The ENPA battery is, indeed, based on a semantic-lexical model³⁶ and included several sections, each with its own cut-off score. Here, the following ENPA tests were administered: word, non-word, and sentence repetition; picture naming of nouns and verbs; auditory word and sentence comprehension; lexical retrieval (fluency) of nouns and verbs. Lexical retrieval was also evaluated using phonemic (initial phoneme: F-L-P), and semantic (categories: car manufacturers, food and animals) fluency tests.³⁷ The syntagma repetition test from the Aachner Aphasie Test (AAT)³⁸ was administered as an additional measure of the repetition ability. Auditory-verbal short-term memory was assessed by means of the Auditory Digit Span³⁹ and a 2-syllable Word Repetition Span.⁴⁰

Lesion data

Brain images were available for 17 out of 22 participants of the PHOLEXSEM group and for all participants of the PACE group. The location and the size of the lesion was reconstructed for each participant and computed by means of a template technique in MRIcro.⁴¹ Figure 1 shows the overlay lesion plot. Mean lesion volume was calculated for each group. See also Table I.

Figure 1.

—Overlay lesion plots for the PHOLEXSEM (N.=17) and PACE (N.=21) groups. Violet areas represent minimum overlap, red areas maximum overlap.

PHOLEXSEM treatment

The experimental PHOLEXSEM protocol focused on the treatment of phonological, lexical, and semantic deficits, with the primary aim of improving lexical retrieval, and, generally, spoken output. Patients underwent 10 daily sessions (2 weeks of treatment, Monday to Friday). Each daily session lasted 30 minutes, namely, three parts lasting 10 minutes with Repetition, SFA, and Lexical retrieval exercises.

Repetition part

Fifty items were presented for each type of stimulus: syllables, 2-syllable words (with Consonant – C – and Vowel – V- sequence: CVCV); 2-syllable words with one consonant cluster (CVC1C2V); 3-syllable words (CVCVCV); 3-syllable words with one consonant cluster (CVC1C2VCV), nuclear phrase (subject-verb - SV); 2-syllable non-words; 2-syllable non-words with consonant cluster; 3-syllable non-words; 3-syllable non-words with consonant cluster. The CV sequences and the clusters were the same for both words and non-words.

If the patient failed to repeat the stimulus at the first presentation, a second presentation was provided. The repetition of each stimulus type was considered successful if the patient made no errors in at least 80% of the items; when this level of accuracy was reached, the subsequent set was administered.

SFA part

The stimuli included 46 pictures of nouns (see the Supplementary Digital Material 1, Supplementary Text File 1 for the list of stimuli) including living (20 items, mean frequency of use=13.9, standard deviation, SD=20.2, range=0-76) and non-living (26 items, mean frequency of use=17.3, SD=23.9, range=0-83) objects, controlled for frequency of use (Banca Dati dell’Italiano Parlato - Institut für Romanistik, Karl-Franzens-Universität Graz, Merangasse 70, A-8010 Graz). Each patient was instructed to name a picture, placed in the center of the feature analysis schema. Regardless of naming accuracy, the patient had also to name the semantic features of each item, that is the category, use, action properties, location, and associations. If the participant gave no response when required to name the picture or report its semantic features, or a paraphasic error was produced, the therapist spoke the unproduced words and wrote them in the boxes, asking the patient to repeat them. If even in this case, after the semantic feature report, the patient was still unable to name the picture, the therapist said the name aloud requiring its repetition. The same stimulus was never repeated during the 10 different rehabilitation sessions. On average, each patient completed 4/5 items.

Lexical retrieval part

Three fluency tasks were given: 4 minutes of phonemic recall, 4 minutes of semantic recall, and 2 minutes of verb recall. No items used for rehabilitation were employed in the evaluation sessions. In particular, for the phonemic recall, in the first eight sessions (from day 1 to day 8) patients were asked to recall words based on the initial phoneme, while in the last two sessions (day 9 and day 10), word recall had to made according to the final vowel. For the semantic recall, different categories were given, varying in concreteness (e.g., vegetables) and abstractness (e.g., feelings). Finally, the verbal recall task required the participants to produce verbs related to a word orally presented by the therapist; for instance, when the therapist said “car,” the related verbs could be “drive, brake, accelerate, etc.”).

In every task, if the patient produced no correct response within 15 seconds, a cue was provided by the therapist. In the phonemic recall task, oral and written cues were presented, i.e., a vowel was added the consonant to form the first syllable of the word to be produced. For instance, for the recall of words starting with “C”, the therapist said “CA” or “CO” and wrote it on a sheet. For semantic and verbal recall, the therapist provided verbal suggestions: for example, for “clothes” the therapist suggested “Think about your wardrobe” or “Clothing according to the season.”

PACE treatment

During the PACE, three decks of cards depicting nouns with different frequency of use were presented in sets of four cards each. In an exchange modality, the task consisted in understanding the noun chosen by the patient and, in turn, by the rehabilitator. Any communicative modality (oral, gestural, verbal, graphic) was allowed. Word difficulty was graded based on the frequency of use and semantic proximity or distance between target and distractors. For details see Carlomagno et al.⁷

Statistical analysis

All analyses were carried out with jamovi 2.5.⁴² Alpha was set 0.05.

Between-group comparisons of demographic (i.e., age, education, and sex) and clinical variables (i.e., duration of disease, fluent vs. non-fluent aphasia type, lesion volume, and FIM score) were performed with independent samples t-tests or non-parametric Mann-Whitney Tests when normality assumptions were violated (in case of continuous variables) or Chi-squared Tests (in case of categorical variables).

Baseline differences (i.e., at Pre 1 and Pre 2) between the experimental PHOLEXSEM group and the control PACE group were assessed for each language test by means of Mann-Whitney Tests. Non-parametric tests were chosen because the assumption of normality (tested with Shapiro-Wilk) was violated for all language outcomes.

We planned to analyze the effects of treatment by means of a mixed Analysis of Variance (ANOVA) with “Timepoint” (i.e., Pre 1, Pre 2, Post 1, Post 2) as within-subject factor and “Group” (i.e., PHOLEXSEM and PACE) as between-subject factor. However, Shapiro-Wilk tests and Q-Q plot inspection indicated that residuals were not normally distributed for any language outcome. Therefore, for each group, non-parametric Friedmann ANOVA was carried out with “Timepoint” as factor. In case of significance, corrected Durbin-Conover post-hocs were calculated.

Finally, for the PHOLEXSEM group, we calculated Spearman correlations between change-scores of each language task (i.e., the difference between Post 2 and the average of Pre 1 and Pre 2 scores) and age, education, lesion volume, disease duration, and the FIM score.

Results

Between-group differences in demographic and clinical variables

No significant differences were found between PHOLEXSEM and PACE groups in terms of age (t₄₁₌-0.702, P=0.487), education (U=207.5, P=0.561), sex (χ²₁₌0.22, P=0.639), days elapsed since stroke onset (t₄₁₌-0.576, P=0.087), lesion volume (U=139, P=0.256), and the frequency of the aphasia type (χ²₁₌0.19, P=0.661). A significant difference was observed only with respect to the FIM score (t₃₃₌6.12, P<0.001): the PHOLEXSEM group had lower functional independence (mean=40.1±20.98 standard deviation) than the PACE group (88.1±25.41).

Treatment outcomes

Mann-Whitney Tests indicated no significant differences at baseline, before the intervention, between the PHOLEXSEM and PACE groups in the neuropsychological assessment of language, both at Pre 1 (highest U =178; lowest P=0.194) and Pre 2 (lowest U=183; lowest P=0.242). Comparisons of all language tests are reported in Supplementary Digital Material 2, Supplementary Table I.

Repeated-measures analyses showed a number of significant improvements for the PHOLEXSEM group. In more details, patients of the PHOLEXSEM group showed a significant improvement of word repetition (χ²₃₌36, P<0.001), non-word repetition (χ²₃₌27.7, P<0.001), sentence repetition (χ²₃₌14.4, P=0.002), as well as an increased repetition score at the AAT (χ²₃₌34.8, P<0.001). Overall, a better performance was achieved at the end of the treatment (i.e., Post 2) compared to both baseline measures, namely Pre 1 (all P<0.001) and Pre 2 (all P<0.008). Relevant comparisons are depicted in Figure 2; see also Table II and Supplementary Digital Material 3, Supplementary Table II for all post-hoc comparisons.

Figure 2.

—Results of repetition tasks. Patient’s performance at the repetition tests (mean score and standard error) are depicted for the PHOLEXSEM (orange lines) and PACE group (blue lines). Asterisks highlight the post-treatment improvements in the PHOLEXSEM group. AAT: Aachner Aphasie Test. *P<0.05; **P<0.01; ***P<0.001.

Table II. —Results of language assessment.

Test	Group	Pre 1	Pre 2	Post 1	Post 2	ANOVA χ2	P
Rep Words	PHOLEX	22.5 [0-30]	21.5 [2-30]	24 [1-30]	28 [1-30]	36*	<0.001*
	PACE	13 [0-30]	15 [0-30]	16 [0-30]	16 [0-30]	00.06	0.996
Rep NW	PHOLEX	9 [0-15]	8 [0-15]	10 [0-15]	12 [1-15]	270.7*	<0.001*
	PACE	4 [0-14]	5 [0-15]	5 [0-14]	4 [0-14]	40.21	0.239
Rep Sent	PHOLEX	3 [0-8]	20.5 [0-9]	4 [0-9]	4 [0-9]	140.4*	0.002*
	PACE	1 [0-8]	1 [0-8]	2 [0-9]	2 [0-8]	40.54	0.208
Rep AAT	PHOLEX	140.5 [0-27]	14 [0-28]	170.5 [0-29]	18 [0-28]	340.8*	<0.001*
	PACE	7 [0-28]	6 [0-28]	5 [0-28]	7 [0-28]	00.86	0.835
Nam Nouns	PHOLEX	1 [0-7]	1 [0-9]	2 [0-10]	30.5 [0-10]	260.6*	<0.001*
	PACE	0 [0-9]	0 [0-9]	1 [0-9]	1 [0-8]	30.38	0.337
Nam Verbs	PHOLEX	1 [0-10]	10.5 [0-10]	2 [0-10]	4 [0-10]	270.5*	<0.001*
	PACE	2 [0-8]	1 [0-7]	1 [0-8]	1 [0-8]	30.57	0.311
Flu Phon	PHOLEX	00.5 [0-9]	1 [0-11]	2 [0-9]	20.5 [0-13]	210.9*	<0.001*
	PACE	0 [0-14]	1 [0-11]	1 [0-12]	0 [0-28]	00.30	0.960
Flu Sem	PHOLEX	1 [0-18]	10.5 [0-24]	6 [0-27]	8 [0-30]	28*	<0.001*
	PACE	1 [0-37]	1 [0-35]	1 [0-37]	1 [0-43]	70.62	0.054
Flu Nouns	PHOLEX	1 [0-7]	00.5 [0-8]	2 [0-10]	3 [0-14]	260.6*	<0.001*
	PACE	0 [0-18]	1 [0-20]	1 [0-17]	1 [0-18]	80.94*	0.030*
Flu Verbs	PHOLEX	0 [0-6]	0 [0-5]	1 [0-8]	3 [0-10]	250.3*	<0.001*
	PACE	0 [0-7]	0 [0-9]	0 [0-11]	0 [0-10]	60.39	0.094
Comp Words	PHOLEX	16 [6-20]	17 [10-20]	18 [8-20]	180.5 [7-20]	50.65	0.130
	PACE	16 [2-20]	17 [2-20]	16 [2-20]	17 [2-20]	50.78	0.123
Compr Sent	PHOLEX	10 [5-14]	9 [4-14]	10 [4-14]	11 [6-14]	90.61*	0.022*
	PACE	10 [6-14]	10 [7-14]	10 [6-14]	10 [7-14]	10.75	0.625
Digit Span	PHOLEX	2 [0-4]	3 [0-5]	3 [0-5]	3 [0-4]	150.9*	0.001*
	PACE	2 [0-5]	2 [0-5]	2 [0-5]	2 [0-5]	20.79	0.425
Word Span	PHOLEX	1 [0-3]	1 [0-3]	2 [0-4]	2 [0-4]	110.3*	0.010*
	PACE	1 [0-3]	1 [0-3]	1 [0-3]	1 [0-4]	100.1*	0.018*

For each language task and timepoint, median and range values are reported. PHOLEX: PHOLEXSEM; Rep: repetition; NW: non-words; Sent: sentences; AAT: Aachner Aphasie Test; Nam: naming; Flu: fluency; Phon: phonological; Sem: semantic; Comp: comprehension. *Significant results of Friedmann’s Analysis of Variance (ANOVA).

Likewise, significant improvements were observed in all tasks involving lexical access, namely, naming of nouns (χ²₃₌26.6, P<0.001) and verbs (χ²₃₌27.5, P<0.001), phonological (χ²₃₌21.9, P<0.001) and semantic fluency (χ²₃₌28, P<0.001), as well as noun (χ²₃₌26.6, P<0.001) and verb fluency (χ²₃₌25.3, P<0.001). The greatest changes emerged at Post 2, as compared to both Pre 1 and Pre 2 (all P<0.001) (Figure 3).

Figure 3.

—Results of naming and fluency tasks. Results of naming and fluency tests (mean score and standard error) are depicted for the PHOLEXSEM (orange lines) and PACE group (blue lines). Asterisks highlight the main post-treatment improvements observed in the PHOLEXSEM group. AAT: Aachner Aphasie Test. *P<0.05; **P<0.01; ***P<0.001.

In terms of comprehension, the PHOLEXSEM group showed significant improvements in the Token Test (χ²₃₌11.8, P=0.008) and the sentence comprehension task (χ²₃₌9.61, P=0.022) from Pre 1 to Post 2 (P<0.001), but no significant difference in the word comprehension (χ²₃ =5.65, P=0.130). Finally, treatment-induced improvements were also observed in the digit span (χ²₃=15.9, P=0.001) and the word span (χ²₃=11.3, P=0.010) tasks.

As for the PACE group, significant improvements were limited to noun fluency (χ²₃=8.94, P=0.030; Post 2 vs. Pre 1: P=0.010) and the word span task (χ²₃=10.1, P=0.018; Post 2 vs. Pre 2, P=0.003). No other significant effects emerged (highest χ²=7.62, lowest P=0.054). See Supplementary Digital Material 4, Supplementary Table III for post-hoc comparisons of the PACE group.

Correlations of language improvements of the PHOLEXSEM group with demographic and clinical variables

Spearman correlations showed a negative association between education and the improvements of word (rs=-0.68, P<0.001) and non-word repetition (rs=-0.45, P=0.037): patients in the PHOLEXSEM group with a lower level of education showed the greatest improvement.

Age correlated negatively with improvements in digit span (rs=-0.59, P=0.004), while the lesion volume was associated to improvements in verb fluency (rs=-0.53, P=0.030): the larger the lesion, the smaller the clinical gain. Full results can be found in Supplementary Digital Material 5, Supplementary Table IV.

Discussion

This study evaluated the efficacy of a new treatment for persons with aphasia, the PHOLEXSEM treatment, an effective rehabilitation intervention for aphasia in the subacute or chronic phase, regardless of the specific language impairment and type of aphasia. The PHOLEXSEM treatment sought to improve phonological and lexical deficits, focusing on anomia, a central deficit of aphasia, through speech production tasks.

The results show improvements in several tasks involving repetition, lexical retrieval and sentence comprehension, which were assessed by administering items not used during the treatment, hence suggesting a generalization effect.⁴³ However, in a future perspective, it would be useful to introduce more ecological improvement measures, such as the evaluation of the discourse and the spontaneous speech, to assess PHOLEXSEM effectiveness in daily life.

Interestingly, the findings from the correlational analyses show that being younger and having more circumscribed brain lesions are associated with larger clinical gains, in line with recent findings (e.g. Kristinsson et al.⁴⁴ for the predictive effect of age), also confirming the impact of the extent of brain damage on language recovery.^{45, 46} Nonetheless, the overall marginal influence of demographic and clinical factors suggests a wide applicability of the PHOLEXSEM for aphasic patients with different disease chronicity, aphasia severity, and educational level. With respect to the latter, the analyses indicate that patients with a lower level of schooling showed even greater improvements in word and non-word repetition.

Compared to the PACE treatment, which follows a communication-based approach, the PHOLEXSEM appears to be effective in only two weeks of treatment. Notably, its ease of administration and malleability make it a valuable protocol even in the subacute phase of the disease, whereby comorbidity or the presence of other cognitive disorders might make it difficult to administer an in-depth neuropsychological assessment of language, which is necessary to set up a more specific, but usually intensive and cognitively demanding treatment. In fact, a major advantage of the PHOLEXSEM is that it does not require the predominant involvement of executive and praxis functions, which can be burdensome in the early phase of the disease.⁴⁷ In addition, the treatment can also be administered by a caregiver, for example if the patient is temporarily unable to travel to treatment facilities or is on a long waiting list to access the rehabilitation services. The type of material used for the rehabilitation makes the PHOLEXSEM treatment also suitable for digital and online administration, in a remote mode that has proved indispensable in the recent pandemic period.⁴⁸

Although it cannot be considered substitutive of specific treatments, such as morphosyntactic or phonological approaches, PHOLEXSEM treatment can be considered a good rehabilitation option to be used flexibly at different stages of the language rehabilitation process, depending of the needs of the patients. In fact, with only a few hours of exercise, it shows good effectiveness on the phonological and lexical side, allowing time for the rehabilitation of other aspects of the aphasic disorder that require a lot of effort to generalize to the ecological context.⁴⁹ Indeed, here we showed that with ten 30-minute daily sessions, the PHOLEXSEM produced significant, albeit small, linguistic improvements, that are not influenced by the severity and type of aphasia, as well as the duration of the disease. The expectation is that a more intensive intervention (e.g., longer duration of daily sessions), with a greater variety of material to be used during the training, could induce larger and long-lasting clinical gains.

Regarding the last point, a major limitation of the present study is the absence of follow-up assessments, as well as a certain degree of clinical and demographic variability of the sample, including the linguistic performance of the two groups at the baseline. Although these differences were mostly non-significant, we did observe differences in functional independence (i.e., the FIM Score). However, the PHOLEXSEM group had lower functional independence than the control group; consequently, one might have expected more severe linguistic impairments and fewer improvements in this group, but this was not the case.

Conclusions

The 2-week PHOLEXSEM, a speech production training that engages several linguistic skills, can improve different aspects of spoken output (ranging from repetition to naming and lexical fluency) and sentence comprehension. The protocol has two main features suitable for its adoption in the in-patient hospital setting: first, it has short duration (2 weeks), and second, it is suitable for both subacute and chronic stroke patients, regardless of their specific aphasic disorder (fluent vs. non-fluent), unless a more severe global deficit is present with total inability to repeat.

Supplementary Digital Material 1

Supplementary Text File 1

List of the used stimuli in the Semantic Feature Analysis (SFA) phase

Supplementary Digital Material 2

Supplementary Table I

Comparisons of baseline performance between PHOLEXSEM and PACE groups.

Supplementary Digital Material 3

Supplementary Table II

Post-hoc comparisons between timepoints – PHOLEXSEM group.

Supplementary Digital Material 4

Supplementary Table III

Post-hoc comparisons between timepoints – PACE group.

Supplementary Digital Material 5

Supplementary Table IV

Correlations between change scores in language tasks, 3 demographic, and clinical variables of the PHOLEXSEM group.