Progression of Motor Speech Function in Speakers With Primary Progressive Apraxia of Speech

Abstract

Purpose:

Speakers with primary progressive apraxia of speech (PPAOS) have an insidious onset of motor speech planning/programming difficulties. As the disease progresses, the apraxia of speech (AOS) becomes more severe and a co-occurring dysarthria often emerges. Here, longitudinal data from speakers with phonetic- and prosodic-predominant PPAOS are used to characterize the progression of their motor speech impairment, including the development of dysarthria and mutism.

Method:

Data are presented from 52 speakers who had PPAOS at enrollment (i.e., progressive AOS in the absence of aphasia, cognitive, or other neurologic symptoms). Twenty-one had predominantly phonetic features, whereas 31 had primarily prosodic features. All participants underwent a comprehensive motor speech evaluation at their enrollment visit and each annual return visit, with a median of three visits per participant.

Results:

Almost 25% of the speakers with PPAOS presented with dysarthria at their enrollment visit (median disease duration of 3.65 years), whereas more than 70% of them had developed dysarthria by their last visit (median disease duration of 6.85 years). Neither the likelihood to develop dysarthria nor the disease duration at which it was detected differed significantly between the phonetic and prosodic groups. However, muteness emerged sooner for speakers with phonetic-predominant PPAOS; the median disease duration at which they became mute was 1.5 years shorter than for their prosodic counterparts.

Conclusions:

Clinically, these results facilitate more accurate prognostication of motor speech symptoms in speakers with PPAOS, allowing for timely introduction of alternative means of communication. The results also support the differentiation between progressive AOS and dysarthria as distinct motor speech disorders that often co-occur in these individuals.

Motor speech disorders, or disruptions in speech with a neurogenic basis, have traditionally been divided into two categories: apraxia of speech (AOS) and dysarthria (e.g., Duffy, 2020; Utianski & Josephs, 2023). AOS is an impairment of the planning or programming processes associated with speaking. Dysarthria is an impairment in the execution of speech of neuromuscular origin; recognizable patterns in motor speech function are associated with specific types of dysarthria and abnormalities in specific structures or pathways of the nervous system. For example, slow speaking rate, imprecise articulation, and strained vocal quality are features of spastic dysarthria and associated with bilateral upper motor neuron (UMN) damage (e.g., Clark et al., 2014; Duffy, 2020). Dysarthria types can co-occur with one another and with AOS (e.g., following a left middle cerebral artery stroke, a speaker might present with both AOS and unilateral UMN dysarthria). Accurate and timely differential diagnosis of disruptions to motor speech informs localization and neurologic diagnosis, which in turn inform prognosis and management. In neurodegenerative conditions, changes to the severity of the motor speech disorder, or the emergence of another type of motor speech disorder, are important indicators of disease progression (e.g., Duffy et al., 2014).

Differential diagnosis of speech patterns is especially informative for disorders that are isolated to motor speech function (e.g., Basagni et al., 2022), including primary progressive apraxia of speech (PPAOS). Speakers with PPAOS present following an insidious onset of motor speech planning/programming difficulties that become increasingly prominent over time to the extent that some of them lose functional use of their speech (e.g., Duffy et al., 2021; Josephs et al., 2006, 2012, 2021; Utianski & Josephs, 2023). Individuals who present with predominant suprasegmental difficulties (e.g., slow rate, segmentation between and within words) are diagnosed with prosodic-predominant PPAOS, whereas those who present with predominant sound-level articulatory errors (e.g., distorted substitutions and additions, especially for complex, multisyllabic words) are diagnosed with phonetic-predominant PPAOS (e.g., Utianski et al., 2018). If the severity of the two categories of features is similar or cannot be compared (e.g., symptoms are too mild or too severe), the speaker is diagnosed with mixed PPAOS. These three diagnoses are best thought of as occurring on a continuum rather than as distinct entities; for example, a speaker might have phonetic-predominant PPAOS, but then go on to develop severe phonetic and prosodic features and be diagnosed with mixed PPAOS at a later visit (Utianski et al., 2023). To be diagnosed with PPAOS, speakers cannot have more than equivocal evidence of aphasia and cannot meet criteria for a broader neurologic syndrome (Duffy et al., 2021; Josephs et al., 2006, 2012, 2021). They can have co-occurring dysarthria, but it must be equal or less severe than their AOS at the time of diagnosis. Differentiating between the features of progressive AOS and dysarthria is therefore critical for arriving at an accurate diagnosis.

Few studies have characterized how motor speech function changes over time in speakers with PPAOS or considered the potential contribution of dysarthria to the observed changes in functional communication. In their longitudinal investigation, Utianski et al. (2021) found that 52% of speakers with PPAOS had developed dysarthria by their final visit (median disease duration 4 years; see also Duffy et al., 2014, 2021). Acoustically, speakers with dysarthria produced words slower than those without, but dysarthria did not influence the change in speech rate across visits. In a group of individuals who had the nonfluent/agrammatic variant of primary progressive aphasia, including some who met criteria for PPAOS, Illán-Gala et al. (2024) found that more severe dysarthria at baseline predicted a faster disease progression, as measured by the Clinical Dementia Rating sum of boxes. Although Bouvier et al. (2021) did not formally diagnose dysarthria, they also commented on changes in vocal quality (e.g., breathiness and strain), hypernasality, and changes on the oral mechanism exam that were “possibly consistent with dysarthria” (p. 1468) over time in their speakers with progressive AOS. Their sample size was limited, but they judged these changes to be more prominent in speakers with predominant prosodic features. Together, these studies justify the need to consider dysarthria in this population when predicting future speech changes.

The presence and type of dysarthria may also provide information about involvement of neural structures other than those associated with progressive AOS and could relate to the broader neurologic syndrome that often emerges over time. Hypokinetic, unilateral UMN, and spastic dysarthria are the most likely to be associated with PPAOS (Duffy et al., 2014). Hypokinetic dysarthria is typically associated with basal ganglia dysfunction and could be related to the atypical parkinsonism that individuals with PPAOS often develop (e.g., Josephs et al., 2014; Seckin et al., 2020; Whitwell et al., 2017). For example, some of them develop features of progressive supranuclear palsy, which is also most commonly associated with hypokinetic dysarthria (Clark et al., 2021). Speakers with prosodic-predominant PPAOS tend to score worse and decline faster than their phonetic counterparts on the motor examination part of the Unified Parkinson's Disease Rating Scale, potentially suggestive of more Parkinsonian features (Utianski et al., 2018; Whitwell et al., 2017). It is therefore possible that speakers with prosodic-predominant PPAOS are more likely to develop hypokinetic dysarthria. Similar predictions for spastic dysarthria are difficult to make a priori as progressive limb spasticity resulting from UMN damage has been reported in individuals with both corticobasal degeneration (e.g., Hasselblatt et al., 2007) and progressive supranuclear palsy (e.g., Krzosek et al., 2022) at autopsy, which are the pathologies most associated with phonetic- and prosodic-predominant PPAOS, respectively (Josephs et al., 2021).

Taken together, speakers who have PPAOS often have dysarthria, but the diagnostic and prognostic value of that co-occurrence has yet to be explored. The goal of this study was to investigate how the presence and nature of the motor speech impairment associated with PPAOS changes over time and whether those changes differ between speakers with prosodic- versus phonetic-predominant AOS. In addition to determining the onset and type of dysarthria in the two groups, we aimed to evaluate time to loss of functional speech to help project when the need for alternative means of communication might be indicated.

Method

Participants

Longitudinal data were reviewed from 56 speakers who were diagnosed with PPAOS at their enrollment visit with the Neurodegenerative Research Group (NRG) at Mayo Clinic. Diagnosis criteria and subtyping are outlined in detail in the next section. Speakers were assigned to the phonetic (n = 21) or prosodic (n = 31) groups based on the consensus diagnosis at their enrollment visit, unless that was mixed. Those who had mixed PPAOS at their enrollment visit were categorized as phonetic or prosodic if that subtype classification became evident at their subsequent visits. Speakers with a mixed classification at all study visits (n = 4) were excluded. Data from all NRG research visits were included for the remaining 52 speakers. There was a median of three visits per participant (range: 1–8) with a total of 167 visits included. Participant and care partner's retrospective report of symptom (i.e., AOS) onset was used to calculate disease duration for all analyses. Written informed consent was obtained from all participants at each visit in accordance with the Mayo Clinic institutional review board (#09–008772, #12–008988, #15–004618, #17–002468). Note that there is overlap between the individuals in this cohort and those included in previous publications by our group (e.g., Duffy et al., 2021; Utianski et al., 2021).

Motor Speech Tasks and Diagnostic Process

Motor speech diagnoses were based on a comprehensive evaluation conducted by a certified speech-language pathologist that included a brief case history and conversation, description of the picnic scene picture from the Western Aphasia Battery–Revised (Kertesz, 2006), an oral motor examination, sustained phonation, alternating and sequential motion rates, and the Articulatory Error Score Index. These speech samples informed ratings on the Apraxia of Speech Rating Scale Version 3 (ASRS-3; Duffy et al., 2023). The ASRS-3 involves clinician ratings of 13 features typically associated with AOS on a scale from 0 to 4, where higher scores indicate increased frequency or severity of the feature. Summed subscores are given for features that reflect phonetic disruption, prosodic disruption, and other aspects of speech, including performance on the alternating and sequential motion rate tasks. Ratings on this measure are made irrespective of the specific speech diagnosis, and some of the rated features (e.g., slow rate) overlap with dysarthria. Note that ASRS-3 scores often cannot be obtained for speakers with severe or profound AOS because several features become impossible to score when there is little to no speech output (21 visits of the 167 reported here). Perceptual ratings on these same tasks were also used to assess the presence, type, and severity of dysarthria. All tasks were video-recorded for later review.

Speech and language diagnoses were verified through consensus from at least three certified speech-language pathologists. In addition to the examining speech-language pathologist, a minimum of two other speech-language pathologists reviewed the audiovisual recordings independently. Each made a diagnosis, and these diagnoses were discussed until agreement was reached by at least two of the three. All participants had unequivocal evidence of AOS and, if they had dysarthria, it was not more severe than the AOS (e.g., Duffy et al., 2021; Josephs et al., 2012). Subtype diagnoses were made following the same consensus protocol based on overall impressions of whether the articulatory or prosodic features predominated. As described above, if this designation could not be made (i.e., because the symptoms were too mild or too severe to judge the relative severity or because the severity was judged to be similar), then speakers were diagnosed with mixed PPAOS.

Subtype judgments were based on an overall impression from all speech tasks. It may be tempting to associate the ASRS-3 subscores with this judgment, but it is important to recognize that most of the ASRS-3 ratings are made irrespective of the source of the feature (e.g., AOS, dysarthria, aphasia), whereas the subtype diagnosis is made only based on the features attributable to the AOS, based on clinical judgment. This distinction is especially critical for speakers who also have dysarthria. For example, a high score on the ASRS-3 feature of slow speaking rate, which is categorized as a prosodic feature, could be attributable to dysarthria, AOS, or speaker-imposed compensatory slowing. In that example, the entity of slow rate would contribute to the prosodic subscore on the ASRS-3 but would only be considered in the AOS subtype determination if AOS was judged to be the source.

To be diagnosed with dysarthria, speakers had to have unequivocal evidence of speech abnormalities that could not be accounted for by the AOS. AOS is only associated with changes to articulation and suprasegmentals, so any consistent abnormalities related to respiration, voice, or resonance point to a co-occurring diagnosis. Unilateral UMN and spastic dysarthria were considered present if consistent articulatory imprecision occurred in the presence of clearly asymmetric or bilateral central facial and tongue weakness, respectively. Bilateral velopharyngeal weakness can cause hypernasality, and the vocal quality can range from harsh to strained. Confirmatory signs include a hypersensitive gag reflex, positive pathological reflexes (e.g., suck, snout, palmomental), and pseudobulbar affect. Changes in vocal quality were also among the most prominent and easily detectable features for the speakers with hypokinetic dysarthria. In this case, the changes often include elevated pitch, breathy-harsh voice, and/or reduced loudness; consistent articulatory imprecision is often also present in the context of reduced range of movement of the articulators during speech, with hypomimia as a confirmatory sign. Even though fast speaking rate is a common feature of hypokinetic dysarthria in isolation, it is our experience that it is rarely observed in speakers who have PPAOS. Rather, their motor speech planning/programming difficulties and rigidity are likely to contribute to an overall slower speaking rate. Hyperkinetic dysarthria is caused by adventitious movements that impact speech. Dysarthria type diagnoses reported in the current study were based on the first visit at which the speaker was found to have dysarthria.

Finally, the classification of mutism was made retrospectively based on a rating of the severity of the motor speech symptoms and associated communication disorder, where 10 reflects normal speech and 0 reflects limited vocalization that is rarely attempted (adapted from Hillel & Miller, 1989; Yorkston et al., 1993). We used a cutoff of 3 or less on this rating scale, where 3 reflects that the individual can only vocalize single word responses and either writes or uses a spokesperson to communicate more complex utterances, initiating communication nonvocally.

All speech and language consensus diagnoses (i.e., AOS, dysarthria, aphasia) were made blinded to neuropsychological, neurologic, and neuroimaging data. Two board-certified behavioral neurologists then established the final neurologic diagnosis. All participants met the exclusionary criteria for PPAOS at the first visit (i.e., did not have aphasia, widespread motor issues, or a gaze palsy), but diagnoses at subsequent visits (e.g., progressive supranuclear palsy, behavioral variant frontotemporal dementia) were obtained following the same protocol and often reflected progression of the disease.

Statistical Tests

Group Comparisons

The comparisons between groups at single time points were based on Wilcoxon rank sum tests for continuous variables and chi-squared tests of independence for categorical variables. For the longitudinal ASRS-3 data, we fit linear mixed effects models to quantify group-wise differences in and rates of change for each subscore using time from AOS onset, or disease duration, as the timescale. The models included the group-by-time interaction as a fixed effect along with participant specific random intercepts and slopes.

Incidence Rates

To compare incidence rates of dysarthria between groups, we used a Kaplan–Meier survival curve and a delayed entry Cox proportional hazard regression model with disease duration as the timescale. To avoid immortal time bias, the starting time was the participant's disease duration at their first NRG visit and their stopping time was the disease duration at the visit dysarthria was first identified (constituting an event) or their last NRG visit (censored). In this way, the start time to stop time interval represents the interval at which the participant was at risk for an observed event (see Therneau & Grambsh, 2000). For example, a speaker who presented for their first visit with a disease duration of 1.3 years and for their final visit with a disease duration of 5.4 years would be included in the model for 1.3–5.4 years post-AOS onset.

Individuals who presented with dysarthria at their enrollment visit (n = 12, eight prosodic and four phonetic) were excluded from these models, because we do not have precise information about the timing of onset of dysarthria (i.e., their event was not observed), as were those without dysarthria who only had one NRG visit (n = 9, six prosodic and three phonetic) because their start and end points were the same. The same approach was used to compare incidence rates of mutism, except that speakers who presented with dysarthria at their enrollment visit were included in this analysis. All analyses were performed using R (Version 4.3.2) with the lme4 and survival packages.

Results

Demographics and speech diagnoses for each group are outlined in Table 1. The two groups did not significantly differ in terms of biological sex or disease duration at enrollment. In contrast, the prosodic group was significantly older, as previously reported (e.g., Utianski et al., 2018). An overview of the progression of motor speech symptoms for each individual is outlined in Figure 1 by PPAOS subtype.

Table 1.

Overview of demographics and group differences.

Demographics and clinical variables	Phonetic (n = 21)	Prosodic (n = 31)	p value
Age at onset	61 (43, 75)	72 (50, 82)	< .001
# Female	10 (48%)	20 (65%)	.355
Time onset to enrollment visit	3.3 (1.7, 10.9)	3.7 (1.7, 7.5)	.918
# of NRG visits	3 (1, 6)	3 (1, 8)	.503
ASRS-3
Phonetic subscore	7 (3, 15)	3 (0, 8)	< .001
Prosodic subscore	4 (0, 11)	8 (4, 12)	< .001
Other subscore	4 (2, 9)	4 (1, 10)	.777
Dysarthria
Present at enrollment	4 (19%)	8 (26%)	.816
Developed by last visit	14 (67%)	23 (74%)	.783
Disease duration when it developed	4.6 (3.3, 13.8)	5.9 (2.2, 11.0)	.888
Mutism
Developed by last visit	11 (52%)	10 (32%)	.245
Disease duration when it developed	6.2 (4.3, 12.6)	7.7 (4.1, 10.9)	.778
Current status (% of subtype group)			.815
Enrolled	4 (19%)	7 (23%)
Deceased	14 (67%)	18 (58%)
Lost to follow-up	3 (14%)	6 (19%)

Note. Times and age are given in years. Values indicate median (minimum, maximum) or n (% of subgroup). Disease duration to when dysarthria developed includes the disease duration at the first visit as a proxy for the speakers who had dysarthria at study enrollment. ASRS-3 subscores are from the first NRG visit. Participants were considered to currently be enrolled if their last visit was within the last year and they are still alive. Bold text indicates significant results at the p < .05 level. NRG = Neurodegenerative Research Group; ASRS-3 = Apraxia of Speech Rating Scale Version 3.

Figure 1.

Progression of motor speech diagnoses. Swimmer plots illustrate the timeline of motor speech diagnoses for each speaker. The x-axis reflects time in years, with 0 being the estimated onset of the apraxia of speech. All subsequent diagnosis changes were made at a Neurodegenerative Research Group (NRG) visit. An X marks the date of death; for speakers who are still alive, the line ends at their most recent NRG visit. The vertical lines indicate median onset of dysarthria and mutism for the respective groups (see also Table 1). AOS = apraxia of speech.

AOS

As expected, at their enrollment visit, the phonetic group had a higher ASRS-3 Phonetic subscore, the prosodic group had a higher ASRS-3 Prosodic subscore, and they did not significantly differ on the Other subscore (see Table 1). AOS severity increased over time, as indexed by each of the three ASRS-3 subscores (see Figure 2). Speakers with phonetic-predominant PPAOS maintained higher Phonetic subscores than their prosodic counterparts and vice versa for the Prosodic subscores, but the rates of change over time were similar for both groups across each of the three subscores. Even though the two groups had similar overall subscores for the Other features, the distribution differed slightly between groups. Using the data from the last visit that the ASRS-3 could be scored, the reduced words per breath group feature was rated significantly higher for the prosodic group than the phonetic group, p = .005 (see Figure 3).

Figure 2.

Apraxia of Speech Rating Scale Version 3 (ASRS-3) results. The left panels reflect ASRS-3 subscores for speakers with prosodic (purple) and phonetic (green) primary progressive apraxia of speech as a function of disease duration. Each dot represents a Neurodegenerative Research Group visit; the thin lines connect visits from the same speaker. The thicker lines represent predicted subscores for each group from the respective linear mixed-effects models. The right panels reflect coefficients with 95% confidence intervals for each variable. Significant effects have confidence intervals that do not cross 0. AOS = apraxia of speech.

Figure 3.

Apraxia of Speech Rating Scale Version 3 (ASRS-3) other features. Mean (standard error) ratings for the ASRS-3 features that contribute to the Other subscore for each group at the last visit that could be scored. AMRs = alternating motion rates; SMRs = sequential motion rates; AOS = apraxia of speech.

To evaluate the correspondence between the ASRS-3 ratings and the subtype diagnosis, we calculated the difference between the Phonetic and Prosodic subscores for each individual at each visit (see Figure 4). Individuals with a positive difference score had higher Phonetic subscores and were almost all diagnosed as having phonetic-predominant PPAOS. The reverse was true for the prosodic group.

Figure 4.

Correspondence between subtype and Apraxia of Speech Rating Scale Version 3 (ASRS-3) difference scores. Each dot represents one visit for each individual. The ASRS-3 difference score is the subtraction of prosodic subscore from the phonetic score. A positive ASRS-3 difference score indicates that the phonetic subscore was higher than the prosodic subscore for that visit. A negative ASRS-3 difference score indicates that the prosodic subscore was higher than the phonetic subscore. The outline of the dots corresponds to the overall subtype diagnosis group. Transparent dots correspond to visits at which the speaker was diagnosed with mixed primary progressive apraxia of speech (PPAOS). By their final visits, 19 of the speakers (12 prosodic, seven phonetic) were judged to have mixed PPAOS due to overall severity. AOS = apraxia of speech.

Dysarthria and Mutism

By their final NRG visit, 37 of the 52 speakers with PPAOS (71%) had developed dysarthria. Among them, dysarthria developed at a median of 5.9 years following symptom onset (see Table 1). The incidence rate of developing dysarthria increased over time at a slower rate for speakers who had prosodic PPAOS relative to those who had phonetic PPAOS; however, this difference was not significant (see Figure 5 and Table 2). Of the speakers who had dysarthria, the distribution of dysarthria type did not significantly differ as a function of AOS subtype (see Figure 6), p = .192. Overall, 21 of the 52 speakers (52% of speakers with phonetic-predominant PPAOS, 32% of speakers with prosodic-predominant PPAOS) had become mute by their final visit; the median disease duration for the development of mutism was 6.2 years for the phonetic group and 7.7 years for the prosodic group (see Table 1). Individuals with prosodic PPAOS had approximately a 70% slower incidence rate of becoming mute than their phonetic counterparts (see Figure 5 and Table 2). Of the 21 speakers who became mute, 14 of them had been diagnosed with dysarthria at a previous visit and seven of them (five phonetic and two prosodic) had not.

Figure 5.

Kaplan–Meier survival curves. Incidence rates of developing dysarthria (top) and becoming mute (bottom) for each group. AOS = apraxia of speech.

Table 2.

Results of the Cox proportional hazard models for phonetic versus prosodic progressive apraxia of speech.

Event of interest	Coefficient	Hazard ratio	95% confidence interval	p value
Dysarthria	−0.28	0.75	[0.31, 1.32]	.532
Mutism	−1.21	0.30	[0.12, 0.73]	.008

Note. Hazard ratios less than 1 indicate that the risk is lower for the prosodic group compared to the phonetic group. Bold text indicates significant results at the p <. 05 level.

Figure 6.

Distributions of dysarthria types. Data include only the first visit at which each speaker was diagnosed with dysarthria. The white numbers in the bars indicate the number of individual speakers with that combination of primary progressive apraxia of speech subtype and dysarthria type. PPAOS = primary progressive apraxia of speech; AOS = apraxia of speech; UUMN = unilateral upper motor neuron dysarthria.

Discussion

Overall, 71% of the individuals with PPAOS in this cohort developed dysarthria by their final research visit, with a median disease duration of 5.9 years at dysarthria onset, highlighting both the prevalence of motor execution disturbances in this population and the (often) delayed onset relative to their motor planning difficulties. As expected, hypokinetic, spastic, and unilateral UMN dysarthria were the most common dysarthria types between the groups. Although neither the distribution of dysarthria types nor the incidence rates of dysarthria significantly differed between AOS subtype groups in this cohort, the incidence rate of mutism did. By their last visit in the study, 52% of the speakers with phonetic-predominant PPAOS and 32% of the speakers with prosodic-predominant had become mute with onset at median disease durations of 6.2 and 7.7 years, respectively.

With respect to AOS, results yielded evidence about how the ASRS-3 subscores relate to clinically judged subtype diagnoses over time. Even though ASRS-3 ratings and clinical judgments were made independently, they were based on the same speech samples and obviously considered many of the same features. It was therefore reassuring, from a concurrent validity perspective, that nearly all of the speakers in the phonetic group had higher Phonetic subscores than Prosodic subscores and vice versa (see Figure 4). This was the only analysis for which we maintained mixed diagnoses; every speaker who had a mixed designation had Phonetic and Prosodic subscores that were within 3 points of one another, reflecting relatively equal severity. The ASRS-3 also includes an Other subscore that includes AOS features yet to be clearly associated with one subtype or the other. The ratings for most of the Other features were similar between groups, but the speakers with prosodic-predominant PPAOS had higher ratings on the reduced words per breath group feature than their phonetic counterparts. These data could be used to motivate reclassification of this feature in future iterations of the rating scale.

There is an ongoing discussion about the high number of overlapping features between spastic dysarthria and prosodic-predominant PPAOS (e.g., Illán-Gala et al., 2024; Utianski & Josephs, 2023). Despite the similarities, there are also nonoverlapping features that are important to consider in distinguishing between AOS and dysarthria (see also Utianski & Josephs, 2023). Speakers with prosodic-predominant PPAOS also have phonetic features of AOS (e.g., distorted substitutions and additions) that would not be expected in someone with isolated spastic dysarthria. The present results demonstrate that these phonetic features also increase in severity over time, as reflected in the ASRS-3 phonetic subscore. If prompted to speak faster, speakers with prosodic-predominant PPAOS are generally able to, but this will also cause an increase in phonetic errors. In contrast, speakers with spastic dysarthria are restricted in the maximum speed of their movements (speech and nonspeech) due to the hypertonia. There are also a subset of features that are observed in spastic dysarthria, but not prosodic AOS, including changes to vocal quality (e.g., strained–harsh), hypernasality, and consistent articulatory distortions, often in the presence of significant velar, lower face, and tongue weakness. The physical examination (e.g., physiological slowness or weakness) can therefore support accurate differential diagnosis.

Clinically, the differentiation between AOS and dysarthria has important implications. Equipped with the knowledge that dysarthria can co-occur in individuals with PPAOS, clinicians can conduct a careful initial assessment and follow speech features over time. The finding that 12 of the 52 speakers here had developed dysarthria prior to enrolling in the study reinforces that monitoring for dysarthria in speakers with PPAOS should begin early. In related disorders such as progressive supranuclear palsy, there is also evidence to suggest that dysarthria presence and type is related to dysphagia symptoms (e.g., Kwon & Lee, 2019; Müller et al., 2001; Petroi-Bock et al., 2024), which can cause life-threatening complications. Recognizing that someone with PPAOS has developed dysarthria should therefore prompt more active monitoring for dysphagia symptoms. Interventions targeting communication also benefit from accurate identification of the speech features and their etiology. For example, articulation difficulties in the context of AOS could be addressed through script training (e.g., repeatedly practicing functional multisyllabic words), whereas script training would not be expected to improve the articulation difficulties coming from rapid speech rate, weakness, or hypertonia in the context of dysarthria. In sum, reliably identifying the presence and type of dysarthria in the context of PPAOS provides information about the neurophysiological progression of the disease, potentially including the emergence of dysphagia, and provides clues about the most effective treatment options.

Of those speakers who had dysarthria, a majority of those with phonetic-predominant PPAOS had hypokinetic dysarthria and a majority of speakers with prosodic-predominant PPAOS had spastic dysarthria. If these patterns hold and reach statistical significance in larger studies, the relationship could help predict other differences that may emerge between the subtypes, the underlying pathology, and pathological spread. For example, only speakers with phonetic-predominant PPAOS had unilateral UMN dysarthria and its associated confirmatory features (e.g., unilateral facial weakness). This pattern fits with the phonetic group most often having corticobasal degeneration as their primary pathology (e.g., Josephs et al., 2021), because this pathology is associated with corticobasal syndrome and asymmetric findings elsewhere (Armstrong et al., 2013). One of those speakers evolved to have involvement of the contralateral side, as evidenced by a spastic dysarthria at subsequent visits; the other continued to demonstrate asymmetric findings at her most recent visit. More generally, the results presented in Figure 6 only reflect the dysarthria type at the visit at which dysarthria was first diagnosed, but with evolution over time, many of the speakers developed features of both hypokinetic and spastic dysarthrias.

In addition to characterizing the presence and type of dysarthria in speakers with PPAOS, this is the first large-scale study to provide estimates of the timeline to the loss of speech as a primary method of communication, which was documented at a median of 7.6 years postsymptom onset across all speakers. The incidence rate of becoming mute rose faster for speakers with phonetic-predominant PPAOS; over 50% of them had become mute by their final visit and the median disease duration at the visit at which they were first judged to be mute was about 1.5 years sooner than their prosodic counterparts (6.2 and 7.7 years following AOS onset, respectively). This finding is consistent with the finding that acoustic measures (e.g., speaking rate) decline faster over time in speakers with phonetic-predominant PPAOS (e.g., Utianski et al., 2021, again recognizing the overlap between the two cohorts). One possibility is that the more rapid emergence of mutism in the phonetic group reflects a greater impact of phonetic than prosodic features on intelligibility (see, e.g., Lee et al., 2014). In addition, speakers with phonetic-predominant PPAOS had a median time to dysarthria that was 1.5 years shorter. That difference was not statistically significant, potentially owing to the relatively small sample size and within-group variability; however, earlier dysarthria symptoms that progress over time may have also contributed to the earlier loss of functional speech. Even when dysarthria was present, the AOS remained the more severe motor speech disorder over the course of the disease for nearly all speakers and was likely the primary cause of their muteness. Indeed, one-third of the speakers who became mute had not previously been diagnosed with dysarthria, suggesting that AOS alone can lead to muteness (see Figure 1).

Recognizing that there is considerable variability among speakers, such that a subset continued to have functional speech at their time of death, these results provide a useful benchmark for clinicians, speakers with PPAOS, and their families to ground prognosis and planning for alternative communication systems. Unfortunately, we did not collect systematic data about when participants began to use communication methods to supplement speech or which systems they preferred to use. Anecdotally, most of them seemed to prefer lower technology options (e.g., gestures, writing on paper or with an LCD writing tablet, such as a Boogie board), even if they had more sophisticated electronic devices available. At later stages of the disease, poor manual dexterity due to limb apraxia, spasticity, or rigidity can make writing and using devices more problematic. Responding to yes/no questions with a thumbs-up/thumbs-down or head nod/shake also poses problems for those individuals who reverse yes and no (e.g., Meade et al., 2024). Choosing a system together with the speaker and their care partners is critical to ensure that it is a viable option for them. It is also important to introduce the system early so that they have time to practice using it before losing functional speech and ideally prior to the onset of cognitive impairments that make it more difficult to learn the new system (see also, e.g., Friek-Oken, 2008). Ongoing monitoring of each strategy and devices' utility is important for shifting to alternatives, as needed.

These conclusions should be interpreted in the context of limitations that may impact the results. Some of the results seem to reflect clinically informative differences, but were not statistically significant, possibly reflecting the relatively small sample size and variability across clinical courses. This individual variability is often masked in group analyses but central to clinical care. In future longitudinal studies, it will be helpful to identify the degree of change on the ASRS-3 and other measures that exceeds measurement error and reflects a meaningful difference to the clinician, the speaker, and their communication partners (see Stipancic et al., 2018, for an example with sentence intelligibility in people with amyotrophic lateral sclerosis). There were also inherent limitations regarding estimates of timing and incidence. Data included the same speech assessment protocol at each visit to track changes over time at various points in the disease. The accuracy of the analyses is necessarily constrained by the number and timing of assessments, which were typically separated by 1 year, and relies on speaker and caregiver estimates of AOS symptom onset. For example, a fair number of speakers already had dysarthria at enrollment (12 of 52, 23%). Using the disease duration at their initial research visit as a proxy for estimating the median time to dysarthria onset and excluding them from the incident rate analyses likely means that our estimates of when dysarthria develops are slightly late. The available information only included consensus judgments of the presence and type of dysarthria, but tracking the emergence of specific features over time might be an informative future direction. Finally, there were speakers who experienced several years of disease progression following their final visit, often due to physical impairments that precluded their ability to travel, so our estimates of the incidence of dysarthria and mutism are likely underestimates. Nevertheless, the timelines that we documented are clinically meaningful, reinforcing the need for differential diagnosis and timely introduction of alternative means of communication for these individuals.

Data Availability Statement

The data sets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

Funding Statement

The authors are grateful to the participants and their families for making this research possible. Funding was provided by the National Institutes of Health Grants R01-DC14942 (awarded to Utianski/Josephs), R01-NS89757 (awarded to Josephs/Whitwel), R01-DC12519 (awarded to Whitwell), and R01-DC010367 (awarded to Josephs).