Introduction
Medically unexplained symptoms (MUS) are hugely common across the medical practice. They are often chronic, disabling, associated with very high health and social care expenditure, and have major personal and family impact in terms of quality of life and financial security (Creed & Barsky, Reference Creed and Barsky2004). Traditionally the diagnosis of MUS has adopted an exclusionary approach (tests are normal, therefore it is MUS) and pathophysiological understanding has focused on psychological causation, in particular, the idea that physical symptoms are an expression of underlying anxiety. This has informed treatment approaches which rely strongly on reassurance regarding the lack of serious underlying physical illness, the reattribution of physical symptoms to psychological causes, and the psychological and pharmacological treatment of anxiety/depression. The diagnosis is heavily stigmatised with many healthcare professionals viewing such patients as not genuinely ill, alongside general negative societal attitudes to psychological v. physical illnesses.
In contrast, the last 15–20 years have seen a resurgence of scientific, clinical and service development interest in functional neurological disorder (FND) (Espay et al., Reference Espay, Aybek, Carson, Edwards, Goldstein, Hallett and Morgante2018). This work has confirmed FND to be a very common diagnosis in modern neurological practice (about 16% of new neurology outpatient attendances, about 10% of admissions to hyperacute stroke services) (Stone et al., Reference Stone, Carson, Duncan, Roberts, Warlow, Hibberd and Sharpe2010), that it is associated with low rates of misdiagnosis, and that long-term prognosis with regard to disability and quality of life is poor, similar to that seen in multiple sclerosis and Parkinson's disease (Anderson et al., Reference Anderson, Gruber-Baldini, Vaughan, Reich, Fishman, Weiner and Shulman2007; Gendre et al., Reference Gendre, Carle, Mesrati, Hubsch, Mauras, Roze and Garcin2019; Stone, Sharpe, Rothwell, & Warlow, Reference Stone, Sharpe, Rothwell and Warlow2003). Major efforts have been made to change the diagnostic approach from an exclusionary one to a positive one based on specific symptoms and signs, and for this to be reflected in diagnostic explanation (APA, 2013). Rather than suggesting it is ‘unexplained’, the modern diagnosis of FND emphasises that it is a specific diagnosis which has an underlying mechanism. Here much work has been undertaken to provide a neurobiological dimension to pathophysiological explanations (Baizabal-Carvallo, Hallett, & Jankovic, Reference Baizabal-Carvallo, Hallett and Jankovic2019; Edwards, Adams, Brown, Parees, & Friston, Reference Edwards, Adams, Brown, Parees and Friston2012). This does not seek to ignore or downgrade a psychological level explanation, but rather to explain the brain basis of symptoms in addition. There has been a consequent rebalancing of predisposing factors in FND (e.g. past trauma) to consider them as risk factors that may or may not be relevant to symptom development (Ludwig et al., Reference Ludwig, Pasman, Nicholson, Aybek, David, Tuck and Stone2018). This allows a more bespoke approach to diagnostic explanation, formulation and treatment, reflected in the development of specific psychological and physical rehabilitation techniques that do not depend on Freudian notions of repressed trauma and the catharsis of psychoanalytical exploration (Espay et al., Reference Espay, Aybek, Carson, Edwards, Goldstein, Hallett and Morgante2018).
These developments have resulted in somewhat of a disconnect between diagnostic classification and current scientific evidence for those diagnosed with functional neurological symptoms and for those with ‘MUS’ in general. This disconnect reflects a long-standing division in (psychiatric) classification schemes between conversion disorder and somatisation disorder. In the latest iteration of the Diagnostic and Statistical Manual of Mental Illness (DSM 5), Conversion Disorder was moved from the Dissociative disorders category to Somatic symptom disorder category and relabelled as Functional Neurological Symptom Disorder/Conversion disorder. The diagnostic emphasis switched to positive neurological symptoms and signs, and that the diagnosis did not depend on the identification of conflicts or other stressors though it is acknowledged that these might often be present and might be relevant (APA, 2013). However, this diagnosis only covers functional motor symptoms, symptoms of sensory loss/disturbance (but not pain), and non-epileptic attacks. This restrictive definition is in direct opposition to the very common presence of non-motor symptoms in those with functional motor symptoms, in particular pain, fatigue and cognitive symptoms such as cognitive ‘fog’. In previous work by ourselves and others, such symptoms in addition to depression and anxiety correlated with health-related quality of life (HRQoL), but not with an objective rating of motor symptom severity (Vechetova et al., Reference Vechetova, Slovak, Kemlink, Hanzlikova, Dusek, Nikolai and Serranova2018). Neurobiological models for the FND are in fact agnostic to the nature of the symptom – the same underlying mechanism can account for motor, sensory, cognitive and interoceptive phenomena (Edwards et al., Reference Edwards, Adams, Brown, Parees and Friston2012; Van den Bergh, Witthoft, Petersen, & Brown, Reference Van den Bergh, Witthoft, Petersen and Brown2017). Despite this clinical and scientific background, pain, fatigue and other symptoms in people with FND are currently classified separately in DSM-5, for example as somatic symptom disorder (e.g. with predominant pain), but only if psychological distress regarding symptoms is judged to be ‘excessive’, or with another label such as chronic pain syndrome (APA, 2013). A similar diagnostic division is present in the International Statistical Classification of Diseases and Related Health (ICD)-10 where there is one diagnostic category for the dissociative motor disorder (F44.4) and another for persistent somatoform pain disorder (F45.4) (WHO, 2018).
Here we sought to provide evidence that might shed light on this complex and unsatisfactory situation. We performed a detailed clinical assessment of symptoms, quality of life and disability in a large cohort of patients with a motor FND. We specifically wished to determine the presence and nature of correlations between specific symptoms (motor, non-motor, psychological) and quality of life/disability. Also, we wished to determine if there were specific clusters of patients based on specific symptoms, supporting the current symptom-based diagnostic classification schemes.
Materials and methods
One hundred and ninety-five consecutive patients diagnosed with clinically definite motor FND according to Gupta and Lang criteria [141 females, mean age 46.3 (standard deviation, s.d. = 12.1, range 19–81) years; mean disease duration: 7.3 (SD 7.0) years] in the specialised outpatient service for motor FND at the Neurology Department of Charles University in Prague, 1st Faculty of Medicine and General University Hospital (Gupta & Lang, Reference Gupta and Lang2009) from January 2017 to March 2020 (until the beginning of the coronavirus pandemic) were included in the study. Patients who visited after the beginning of the coronavirus pandemic (i.e. from 4/2020 later) were not included as there could be multiple biases.
Exclusion criteria included age <18 years old, MRI abnormality, intellectual disability, major neurological conditions affecting the central nervous system and/or interfering with motor function (e.g. Parkinson's disease, multiple sclerosis, stroke), psychotic spectrum disorders, bipolar disorder and substance use disorder.
The diagnosis of motor FND was based on detailed clinical interviews and examination by an experienced movement disorders specialist based on positive signs of functional weakness or abnormal movements inconsistent and incongruent with known movement disorders (Espay et al., Reference Espay, Aybek, Carson, Edwards, Goldstein, Hallett and Morgante2018; Gupta & Lang, Reference Gupta and Lang2009). The study was approved by the local ethics committee and all participants gave their written consent to take part in the study.
Objective assessment of motor symptoms
The motor symptoms were classified as functional weakness, tremor, dystonia, myoclonus, gait disorder, or speech disorder.
Dominant (most severe and/or most frequent motor symptom) and additional motor symptom types (i.e. tremor, dystonia, gait disorder, myoclonus and weakness) were identified and the number of different motor symptoms in each patient was used as a proxy measure for motor disorder complexity.
The severity of the motor disorder was assessed using The Simplified FMD Rating Scale (S-FMDRS) (Nielsen et al., Reference Nielsen, Ricciardi, Meppelink, Holt, Teodoro and Edwards2017). The presence or absence of abnormal movement at each of seven body regions (face and tongue, head and neck, left upper limb and shoulder girdle, right upper limb and shoulder girdle, trunk and abdomen, left lower limb, right lower limb) was recorded and rated according to symptom severity and duration (maximum score: 54).
Gait aid score (10 m minimal distance) was evaluated as normal gait = 0, abnormal gait no need for assistance or walking aids = 1, assistance or walker or crutches needed = 2, wheelchair dependent = 3). The criteria for classifying patients as wheelchair dependent were based on the objective gait assessment and only those patients who were completely unable to walk (with or without assistance/support) were classified as wheelchair dependent. Patients using a wheelchair for transportation (some of them for excessive pain, fatigue or low tolerance of exercise rather than motor disorder) but able to walk a short distance (10 m) during the examination were assigned to other groups.
Objective assessment of gait function (S-FMDRS gait subscore = sum of severity and duration of gait disorder, range 0–6) was also used for analysis (Nielsen et al., Reference Nielsen, Ricciardi, Meppelink, Holt, Teodoro and Edwards2017). The presence of instability during the neurological examination was recorded (present = 0, absent = 1). Postural instability was classified as present if the patient was not able to stand/walk without support. Positive functional Romberg or pull test backwards were also considered a sign of postural instability. History of falls or instability was not taken into account.
Subjective assessment of motor and non-motor symptoms
All patients evaluated their own motor symptom severity on a 3-point Likert scale (not bothered at all = 0, bothered a little = 1, bothered a lot = 2) according to the Patient-Health-Questionnaire (PHQ-15). The scale considered 5 motor symptoms categories. In addition to PHQ-15 items assessing motor function including weakness (1), motor coordination impairment (2) and gait disorder (3), we added one item assessing tremor and jerks, i.e. merging tremor and myoclonus together (4) and one item assessing abnormal postures or spasms (5). The total score (subjective motor symptoms severity, SMSS, range 0–10) was calculated.
Additionally, all patients completed questionnaires for depression, anxiety, fatigue, pain, cognitive complaints and HRQoL.
To measure depressive symptomatology the Beck Depression Inventory (BDI-II) was used, consisting of 21 items with a total score 0–63 (Beck, Ward, Mendelson, Mock, & Erbaugh, Reference Beck, Ward, Mendelson, Mock and Erbaugh1961).
To measure levels of anxiety we used the State-Trait Anxiety Inventory (STAI X-1, STAI X-2), a measure of state (20 item STAI X-1) and trait anxiety (20 items STAI X-2) with the range 20–80 for each part (Spielberger, Reference Spielberger1983).
Fatigue was assessed using the Fatigue Severity Scale (FSS), a 9-item scale with the range 1–7 focusing on a functional impact and severity of physical and mental fatigue (Krupp, LaRocca, Muir-Nash, & Steinberg, Reference Krupp, LaRocca, Muir-Nash and Steinberg1989).
To assess pain, we used the PainDetect visual analogue scale (VAS) with the range 0–10 for each subscale (VAS, 0 = no pain, 10 = maximum pain) scales for evaluation of current pain intensity, the average pain and the maximal pain in last 4 weeks. The average of these values (the current, the average and the maximal pain intensity = Pain composite score, total score 0–30) for each subject was used for analyses (Freynhagen, Baron, Gockel, & Tolle, Reference Freynhagen, Baron, Gockel and Tolle2006).
Subjective cognitive complaints were measured using the Czech validated version of the Cognitive Complaints Questionnaire (Le questionnaire de plainte cognitive, QPC), based on an original French 10- item dichotomous (yes/no) questionnaire assessing the presence of cognitive difficulties in the last 6 months with the range 0–10 (Markova et al., Reference Markova, Andel, Stepankova, Kopecek, Nikolai, Hort and Vyhnalek2017). The first two items inquire about general memory abilities, while the remaining eight items inquire about more particular cognitive complaints including difficulties with spatial orientation, language, instrumental activities and personality change.
HRQoL was assessed using the 12-Item Short-Form Health Survey (SF-12) (Ware, Kosinski, & Keller, Reference Ware, Kosinski and Keller1996). Physical Functioning, Role Limitations (both Physical and Emotional), Social Functioning, Pain, Mental Health, Vitality and General Health are domains of HRQoL that are reflected in SF-12 (total score 12–44, higher scores associated with better HRQoL). In order to control for possible autocorrelation bias from the partial overlap of several SF-12 items with measures of anxiety, depression, fatigue and pain we calculated the SF-12 general health subscore including only items regarding the impact of general health state (i.e. SF-12 items 1, 2, 3, 4, 5, 9, 12; total score 7–25) while excluding items related to mental health, mood and emotional problems, bodily pain and fatigue.
To measure a health state to complement the HRQoL, the EuroQoL 5-dimension 3-level instrument (EQ-5D-3L) descriptive part (EQ-5D, range 5–15) and visual analogue scale (EQ-VAS, range 0–100%, with 100% being the best imaginable state of health) were used. Five dimensions are reflected in EQ-5D: mobility, self-care, usual activities, pain/discomfort and anxiety/depression, with three response categories each (no problems, some problems and severe problems) (Rabin, Gudex, Selai, & Herdman, Reference Rabin, Gudex, Selai and Herdman2014).
Statistical analysis
Pearson's correlation coefficient was computed to explore the bivariate relations between variables. Lasso regression with 10-fold cross-validation was used to identify variables affecting the HRQoL measures to later enter a multiple linear model (Friedman et al. Reference Friedman, Hastie and Tibshirani2010). Candidate covariates entering the Lasso model were: age, sex, disease duration, subjective motor symptoms severity, S-FMDRS total score, motor phenotype complexity (number of motor symptoms), S-FMDRS gait subscore, presence of gait abnormality and instability, gait aid score, STAI X-2, BDI-II, QPC, FSS and Pain composite score.
Complete hierarchical clustering using Euclidean distance was used to find putative clusters in data. In particular, we aimed to identify subgroups of patients, where patients in one group had similar characteristics, but different from the patients in other groups. We considered three sets of data when finding clustering: (i) all variables entering the Lasso model, (ii) all variables entering the Lasso model plus the indicators of primary and secondary motor symptoms, and (iii) non-motor variables only (STAI X-1 and STAI X-2, BDI- II, QPC, FSS, and Pain composite score). The data were standardised using the z-score transformation to balance the influence of individual variables, whose original scales could differ by an order of magnitude. Highly correlated variables of STAI X-1, STAI X-2, and BDI-II were decorrelated (replaced by principal components). The significance of putative clustering found was assessed using the gap statistics (Tibshirani, Walther, & Hastie, Reference Tibshirani, Walther and Hastie2001).
Statistical analyses were carried out in R (R Core Team, 2020) using glmnet package for Lasso modelling (Friedman et al., Reference Friedman, Hastie and Tibshirani2010), cluster package for gap statistics calculation (Maechler, Rousseeuw, Struyf, Hubert, & Hornik, Reference Maechler, Rousseeuw, Struyf, Hubert and Hornik2021), and idendro package for interactive dendrogram exploration (Sieger, Hurley, Fišer, & Beleites, Reference Sieger, Hurley, Fišer and Beleites2017). Corrections for multiple testing were intentionally not performed in order to enable inspection of raw p values, e.g. those of correlations between selected pairs of variables of interest.
Results
All consecutive 195 patients with motor FND fulfilling inclusion criteria underwent a full clinical assessment and agreed to fill the questionnaires, however, 17 patients did not return the questionnaires and 26 patients did not complete all questionnaires. All subjects with missing data were excluded from the analysis.
Complete dataset was obtained from 152 patients with clinically definite motor FND (109 females) with mean age 46.0 (SD 12.2) years, mean disease duration was 6.6 years, median 5 years.
Forty-three patients were excluded from the analysis because of missing data [32 females, mean age 47.5 (SD 11.7) years, mean disease duration: 10.0 (SD 7.0) years, median 8 years]. A significantly earlier motor FND onset and longer disease duration (p < 0.001) than in the analysed sample could partially explain lower compliance in this group. In most of these patients, FND had started before a specialised service for FND patients was established in 2015. Chronic course with exposure to numerous diagnostic procedures and a lack of effective treatments might have affected the willingness to collaborate on research. No significant differences were found between the groups in either of the motor domains.
Objective motor symptom characteristics are presented in Table 1.
Table 1. Objective characteristics of motor symptoms - dominant and additional motor phenotype
a Number of patients.
b Numbers give percentages (%) in whom given motor symptom was present as dominant phenotype.
c e.g. 42% of patients with primary gait disorder suffered from secondary tremor.
d Percentages of patients reporting postural instability out of the total number of patients in whom given motor symptom was present as dominant phenotype e.g. 67% of patients with primarily gait disability reported postural instability.
In our cohort, 29% had a monosymptomatic motor presentation, 41% presented with two different types of motor symptoms. Only 3% of patients showed more than 4 phenotypes.
Mean S-FMDRS was 11.3 (SD 8.0, range 0–39). The mean S-FMDRS gait subscore was 2.8 (SD 2.2, range 0–6). Instability during the neurological examination was present in 33% of subjects.
Normal gait was present in 36% of patients, 44% of patients had gait disorder without the need for assistance or walking aids, 16% of patients needed assistance, walker or crutches. Only 4% of patients were wheelchair dependent.
Data from questionnaires on non-motor symptoms, self-reported severity of motor symptoms and HRQoL in patients are shown in Fig. 1.
Fig. 1. Self-reported/subjective measures of motor and non-motor symptom severity and HRQoL. Boxplots and histograms of age, motor and non-motor symptom severity, and HRQoL. Colour dots represent individual patients (n = 152) with their primary motor phenotype. BDI-II = The Beck Depression Inventory II; EQ-5D descriptive part of EQ-5D-3L; EQ-VAS = EQ visual analogue scale, part of EQ-5D-3L; EQ-5D-3L = EuroQoL 5-dimension 3-level instrument; FSS = The Fatigue Severity Scale; Pain = The PainDetect scale items -mean from three values the current/average/maximal pain intensity; QPC = The Cognitive Complaints Questionnaire; SD = standard deviation; SF-12 = The 12-Item Short Form Health Survey; SMSS = subjective motor symptoms severity, STAI X-1/STAI X-2 = The State/Trait Anxiety Inventory.
Correlation analysis
Correlation analysis evaluated the relation between the following domains: age, age of motor FND onset (FMD onset), disease duration, number of motor phenotypes, S-FMDRS total score, S-FMDRS gait subscore, gait aid score, SMSS score and non-motor domains (BDI-II, STAI X-1,2, FSS, QPC and Pain composite score) including HRQoL (SF-12 score, SF-12: general health subscore, EQ-5D, EQ-VAS score).
The main correlation analysis results are shown in Fig. 2, additional/complementary correlation analyses are reported in the following summary of the results. The complete set of correlation analysis results is presented in Online Supplementary Fig. S1.
Fig. 2. Correlations between main objective and subjective domains and SF-12. Bivariate scatter plots and boxplots are shown below the diagonal. Note the absence of diverse clusters in the data. Above the diagonal, there are Pearson's correlations coefficients and their significance shown. Note the high correlations within the block of motor symptoms (green), and within the block of non-motor symptoms (blue) and QoL (yellow). The Subjective motor symptoms severity (SMSS) correlated with all other domains. Each measure (e.g. number of motor phenotypes, S-FMDRS etc) is projected on x-axis beneath its corresponding label on the diagonal and on the y-axis to the left of the label. BDI-II = the Beck Depression Inventory II; FSS = the Fatigue Severity Scale; Gait aid score (0 = normal gait, 1 = abnormal gait no need for assistance or walking aids, 2 = assistance or walker or crutches needed, 3 = wheelchair dependent); Pain = the PainDetect scale items-mean from three values the current/average/maximal pain intensity; QPC = the Cognitive Complaints Questionnaire; SF-12 = the 12-Item Short-Form Health Survey (total score 12–44, higher scores associated with better HRQoL); S-FMDRS = the Simplified FMD Rating Scale (0 – … most severe motor symptoms); SMSS = Subjective motor symptoms severity; STAI X-2 = the State/Trait Anxiety Inventory. * p < 0.05, ** p < 0.01, *** p < 0.001.
Age was positively correlated to subjective cognitive complaints (QPC scores) (p < 0.001), trait anxiety (STAI X-2 score) (p < 0.01) and negatively to the quality of life (SF-12) (p < 0.01), the general health subscore of SF-12 (p < 0.001) and EQ-VAS score (p < 0.01). A weak positive correlation (p < 0.05) was revealed for state anxiety (STAI X-1 score), BDI-II and S-FMDRS gait subscore.
There was found a significant positive correlation between disease duration and fatigue (p < 0.001). Disease duration negatively correlated with gait aid score (p < 0.01), and weakly with S-FMDRS gait subscores (p < 0.05).
All objective measures of motor symptom severity and complexity (number of motor phenotypes, S-FMDRS total score, S-FMDRS gait subscore, gait aid score) correlated with each other (p < 0.001). The S-FMDRS total score significantly correlated with all non-motor symptoms measures (BDI-II, STAI X-1,2, QPC, FSS, pain score). On the other hand, the number of motor phenotypes correlated only with subjective cognitive complaints score (QPC) and EQ5D score (p < 0.001), and weakly with pain and SF-12 scores.
S-FMDRS gait subscore correlated with other objective measures of motor symptom severity (number of motor phenotypes, S-FMDRS total scores) (p < 0.001), but also with all HRQoL measures (p < 0.01) and all non-motor scores (p < 0.05) (Fig. 2).
The subjective motor symptoms severity score significantly correlated with objective measures of motor symptom severity assessed using the S-FMDRS total scores (including S-FMDRS gait subscore, p < 0.001), and all non-motor and QoL scores (p < 0.001) (Fig. 2).
All non-motor measures (BDI-II, STAI X-1,2, FSS, Pain composite score, QPC) correlated strongly with each other and with the SMSS score. The strongest correlation was observed between depression (BDI-II score) and anxiety (STAI X-1,2 score) and cognitive complaints (QPC score).
Both measures of motor symptom severity, the subjective and objective (SMSS, Number of motor phenotypes, S-FMDRS scores, S-FMDRS gait subscores) correlated with HRQoL measures (SF-12 and EQ-5D-3L). SF-12 score and SF-12: general health subscore correlated equally with most measurements.
Although no differences in SF-12 and EQ-5D-3L scores (EQ-5D and EQ-VAS, respectively) were found between patients with dominant gait disorder and patients with other dominant phenotypes (p = 0.63, p = 0.58, respectively), the presence of postural instability was associated with worse scores of SF-12 and EQ-5D-3L (both p < 0.001). Similarly, more severe impairment in gait as measured by the use of walking aids (gait aid score up to the value of 2) was associated with worse scores of SF-12 and EQ-5D-3L (both p < 0.001). Nevertheless, wheelchair dependent patients reported only worse EQ-5D (p < 0.001) and general health subscore of SF-12 (p = 0.01), but not SF-12 (p = 0.19) or EQ-VAS score (p = 0.32) compared to patients without gait problems.
Age of motor FND onset correlated significantly only with S-FMDR gait subscore and gait aid score (shown in the Online Supplementary Fig. S1).
No significant correlations were found between disease duration and SF-12 and EQ-5D-3L scores.
All non-motor measures strongly correlated with HRQoL measures (SF-12 and EQ-5D-3L).
Predictors of HRQoL
Multiple linear regression revealed BDI-II (p < 0.001), Pain composite score (p < 0.001), SMSS score (p = 0.008), STAI-X2 (p = 0.010), and FSS (p = 0.03) were the factors affecting jointly the HRQoL (the SF-12 score).
Similarly, the multiple linear regression model of the subscore of SF-12 related to general health revealed that FSS (p < 0.001), BDI-II (p < 0.001), Pain composite score (p = 0.010), age (p = 0.008) and Subjective motor symptoms severity (p = 0.047) were the factors affecting jointly the HRQoL.
The current health status (EQ-5D measures) was strongly affected by BDI-II scores (p < 0.001), need for use gait aids (Gait aid score) (p < 0.001), acute pain scores (p = 0.002) and S-FMDRS (p = 0.009). The health status measured using EQ-VAS was affected by Pain composite score (p < 0.001), STAI-X2 (p = 0.002), SMSS (p = 0.003) and age (p = 0.003). The effect of S-FMDRS on SF-12 and EQ-VAS was not significant when adjusting for the other factors in the multiple linear model, it only affected the EQ-5D.
Cluster analysis
The cluster analysis revealed that the patients could not be reliably separated into several subgroups: the gap statistic insinuated that the patients formed a relatively homogeneous cluster. This result was found for each of the three data sets considered.
Discussion
Correlation and cluster analyses of self-evaluated and objectively assessed motor symptoms, self-evaluated non-motor symptoms severity and quality of life in a relatively large cohort of patients with heterogeneous motor manifestations including functional weakness provided the following findings.
(1) Objectively assessed motor symptom severity including scales for gait impairment and FND phenotypic complexity correlated with subjectively reported motor symptoms severity. The objectively assessed motor symptom severity using S-FMDRS correlated with all self-reported non-motor symptoms severity scores.
(2) There was a significant mutual correlation between all subjectively reported motor and non-motor symptom measures.
(3) Both the subjective and objective motor symptoms measures showed a significant correlation with HRQoL measures, however, the subjectively reported severity of motor symptoms along with fatigue, pain, depression and anxiety were the main drivers of HRQoL. The objective motor symptoms only partially affected the current health status.
(4) Cluster analysis revealed that the patient sample was relatively homogenous and could not be separated into subgroups based on specific/discrete motor and non-motor features.
These findings suggest that regardless of motor phenotype, there is a continuum in disease severity across multiple domains where patients with mild motor symptom severity reported less severe non-motor symptoms and more severely affected patients reported more severe non-motor symptoms along with worse HRQoL.
Relationship between motor and non-motor symptoms
Consistent with previously reported relationships between multiple non-motor symptoms, (Gelauff et al., Reference Gelauff, Kingma, Kalkman, Bezemer, van Engelen, Stone and Rosmalen2018; Gendre et al., Reference Gendre, Carle, Mesrati, Hubsch, Mauras, Roze and Garcin2019; Vechetova et al., Reference Vechetova, Slovak, Kemlink, Hanzlikova, Dusek, Nikolai and Serranova2018) here we also found relationships between the self-evaluated motor symptom severity and several objective measures of motor impairment. Motor symptom severity assessed using S-FMDRS also correlated with depression, anxiety, fatigue and pain scales. Rather against expectations, no correlation was found between the gait scales and pain.
Interestingly, out of the non-motor symptoms, the subjective cognitive complaint severity was the only measure that correlated with all other subjective and objective motor and non-motor measures which may reflect the role of attentional processes in the development of FND and the importance of the cognitive symptoms (Edwards et al., Reference Edwards, Adams, Brown, Parees and Friston2012; Sadnicka, Daum, Meppelink, Manohar, & Edwards, Reference Sadnicka, Daum, Meppelink, Manohar and Edwards2020; Teodoro, Edwards, & Isaacs, Reference Teodoro, Edwards and Isaacs2018).
The distribution of the data from subjective and objective assessment suggests that patients with objectively less severe motor impairment report having a less subjective motor impairment and less severe non-motor symptoms, i.e. they are not ‘overreporting’ severity of their motor and systematically presenting maximal values.
A significant correlation between objective motor symptom severity and psychological symptom severity (anxiety, depression) has previously been reported in patients with functional myoclonus while it was absent in the organic myoclonus control group (Zutt et al., Reference Zutt, Gelauff, Smit, van Zijl, Stone and Tijssen2017).
Further studies are needed to show whether the pattern of multiple motor and non-motor correlations and a lack of clusters is specific to motor FND or also other FND. Despite the expectation that motor symptoms generally associate with psychological or non-motor symptoms, the literature across different neurological disorders has provided inconsistent results with a large number of studies reporting a lack of correlations in Multiple Sclerosis (Braga, Prado, Bichueti, & Oliveira, Reference Braga, Prado, Bichueti and Oliveira2016) (Arnett, Higginson, Voss, Randolph, & Grandey, Reference Arnett, Higginson, Voss, Randolph and Grandey2002; Bakshi, Reference Bakshi2003; Brassington & Marsh, Reference Brassington and Marsh1998; Krupp, Alvarez, LaRocca, & Scheinberg, Reference Krupp, Alvarez, LaRocca and Scheinberg1988; Krupp et al., Reference Krupp, LaRocca, Muir-Nash and Steinberg1989; Schreurs, de Ridder, & Bensing, Reference Schreurs, de Ridder and Bensing2002; Vercoulen et al., Reference Vercoulen, Hommes, Swanink, Jongen, Fennis, Galama and Bleijenberg1996), Myasthenia gravis (Bartel & Lotz, Reference Bartel and Lotz1995; Chen, Chang, Chiu, & Yeh, Reference Chen, Chang, Chiu and Yeh2011; Doering, Henze, & Schussler, Reference Doering, Henze and Schussler1993; Tennant, Wilby, & Nicholson, Reference Tennant, Wilby and Nicholson1986), adult spinal muscular atrophy (Gunther et al., Reference Gunther, Wurster, Cordts, Koch, Kamm, Petzold and Hermann2019) and Parkinson's disease (Park et al., Reference Park, Youn, Cho, Oh, Kim, Park and Park2018).
Impact of motor and non-motor symptoms on HRQoL
The analysis of the impact of motor and non-motor symptoms on HRQoL revealed a negative correlation between all non-motor scales, motor symptom severity, disability measures and HRQoL measures. Nevertheless, the subjectively reported motor symptom severity rather than S-FMDRS could explain HRQoL, together with depression, pain, anxiety and fatigue. This result extends findings from our previous study conducted in a smaller cohort of motor FND patients which, however, did not consider the self-reported severity of motor symptoms and thus only highlighted the contribution of non-motor symptoms to HRQoL (Vechetova et al., Reference Vechetova, Slovak, Kemlink, Hanzlikova, Dusek, Nikolai and Serranova2018).
The correlation between non-motor measures and HRQoL could result from a significant overlap between the non-motor symptoms measures and several items from the SF-12. To control for this autocorrelation bias between the SF-12 and measures of anxiety, depression, fatigue and pain we performed an analysis with scores only from SF-12 items on general health with the same results.
None of the predominant motor phenotypes was associated with worse HRQoL, nevertheless, patients with the presence of gait impairment (alone or as an accompanying symptom) had worse HRQoL as compared to patients without gait disorder. We also found a relationship between objectively assessed gait severity and the presence of postural instability and impaired HRQoL. These results are similar to those found in disorders such as Parkinson's Disease where postural instability and gait disorder are associated with and impaired HRQoL (Muslimovic et al., Reference Muslimovic, Post, Speelman, Schmand, de Haan and Group2008).
Older age was associated with more severe cognitive impairment and anxiety, more severe gait abnormality and poorer quality of life. Longer disease duration and later disease onset were associated with more severe gait performance and a more frequent need to use gait aids. Interestingly, longer disease duration was not associated with higher non-motor symptoms severity except for fatigue or a higher number of phenotypes (i.e. more complex phenotype).
This pattern is rather against expectations and also differs from most progressive neurodegenerative or neuroinflammatory diseases where long-duration predicts worsening of symptoms and increase in non-motor symptoms frequency and severity across different domains which was documented for example in Motor Neuron Disease (Gunther et al., Reference Gunther, Richter, Sauerbier, Chaudhuri, Martinez-Martin, Storch and Hermann2016) or in Parkinson's Disease (Antonini et al., Reference Antonini, Barone, Marconi, Morgante, Zappulla, Pontieri and Colosimo2012).
Cluster analysis
Patients with motor FND are usually classified according to the dominant motor phenotype they present with (e.g. functional tremor, functional weakness). This is useful when considering differential diagnosis and targeted investigations, and also in physiotherapy management where specific techniques exist for the treatment of specific motor difficulties (Espay & Lang, Reference Espay and Lang2015; Nielsen et al., Reference Nielsen, Stone, Matthews, Brown, Sparkes, Farmer and Edwards2015). Identifying and addressing non-motor symptoms (somatic and psychological) is an additional key part of diagnosis and management (Feinstein, Stergiopoulos, Fine, & Lang, Reference Feinstein, Stergiopoulos, Fine and Lang2001; Garcin et al., Reference Garcin, Mesrati, Hubsch, Mauras, Iliescu, Naccache and Degos2017; Gelauff, Stone, Edwards, & Carson, Reference Gelauff, Stone, Edwards and Carson2014; Jacob, Kaelin, Roach, Ziegler, & LaFaver, Reference Jacob, Kaelin, Roach, Ziegler and LaFaver2018; Maggio et al., Reference Maggio, Ospina, Callahan, Hunt, Stephen and Perez2020; Nielsen, Reference Nielsen2016; Nielsen et al., Reference Nielsen, Stone, Buszewicz, Carson, Goldstein, Holt and Physio2019). We felt it was important, therefore, to analyse whether different combinations of comorbid non-motor symptoms can define more homogeneous/unique subgroups or are associated with specific motor characteristics.
A recent study found no differences in selected characteristics such as demographics, mode of onset and severity of depression, anxiety, pain and fatigue between predefined groups of patients with the different dominant phenotypes (Gelauff, Rosmalen, Gardien, Stone, & Tijssen, Reference Gelauff, Rosmalen, Gardien, Stone and Tijssen2020). Here we used a data-driven approach to search for motor FND subtypes with cluster analysis techniques in an unbiased fashion. Despite a relatively large sample of patients, we failed to identify subtypes based on multiple motor features including motor symptom severity and commonly co-morbid non-motor symptoms in this sample of patients.
In contrast to the lack of clusters in our motor FND group of patients, previous high-quality studies using the same methodology (gap statistics) reported homogeneous clusters including drug-naive parkinsonism (Jain, Park, & Comer, Reference Jain, Park and Comer2015), comorbidities associated with obesity (Reategui, Ratte, Bautista-Valarezo, & Duque, Reference Reategui, Ratte, Bautista-Valarezo and Duque2019), breast cancer progression data (Alexe, Dalgin, Ganesan, Delisi, & Bhanot, Reference Alexe, Dalgin, Ganesan, Delisi and Bhanot2007). However, most cluster analysis studies in neurological conditions with motor symptoms such as Parkinson's disease (Ba, Obaid, Wieler, Camicioli, & Martin, Reference Ba, Obaid, Wieler, Camicioli and Martin2016; Mu et al., Reference Mu, Chaudhuri, Bielza, de Pedro-Cuesta, Larranaga and Martinez-Martin2017; Yang, Kim, Yun, Kim, & Jeon, Reference Yang, Kim, Yun, Kim and Jeon2014) or fibromyalgia (Yim et al., Reference Yim, Lee, Park, Kim, Nah, Lee and Lee2017) suffered from important methodological problems which could have led to false-positive cluster identification. Therefore, making inferences about the specificity of our findings is not possible and further studies are needed.
Interpretation
Our finding of a significant relationship between subjective measures of motor and/or non-motor symptoms and measures of HRQoL may be affected by content overlap across questionnaires. For example, HRQoL questionnaires address the impact of impaired mobility, mood, fatigue on QoL; the BDI scale for depression assessment includes several items on somatic symptoms.
However, the lack of evidence of clusters along with a high correlation between all self-reported measures of motor and non-motor symptoms and HRQoL is entirely consistent with the predictions of predictive coding/active inference accounts of FND. These models suggest that symptoms are perceptions of the state of the body. The symptoms are generated by neural processes that actively sample information from the body and process this information in the context of prior predictions or expectations into conscious perceptions (i.e. symptoms = percepts) (Edwards et al., Reference Edwards, Adams, Brown, Parees and Friston2012; Van den Bergh et al., Reference Van den Bergh, Witthoft, Petersen and Brown2017).
Crucially, these models are agnostic to the content of the percept. It is proposed that in people with FND an abnormal prior expectancy regarding a particular symptom is enhanced in its strength (precision), and this overwhelms incoming sensory data that would indicate a normal state of the body. In this way an abnormal percept results which is experienced spontaneously and involuntarily, without a sense of control or agency over what has been experienced. This same dysfunction can affect motor, interoceptive and exteroceptive control. Therefore, a high degree of cross-correlation could reflect a common dysfunction that underpins motor and non-motor symptoms (Edwards et al., Reference Edwards, Adams, Brown, Parees and Friston2012; Van den Bergh et al., Reference Van den Bergh, Witthoft, Petersen and Brown2017).
This is consistent not only with our data, but also consistent with clinical experience. In patients with functional motor symptoms, multiple somatic symptoms are commonly seen. In some patients the severity of symptoms wax and wane with, for example, the pain becoming more prominent while motor symptoms might improve slightly. Some patients start with chronic pain or fatigue and then later develop functional motor symptoms and vice versa. These phenotypic observations are entirely consistent with a single pathophysiological process which can affect multiple input streams and the sensorimotor control of movement.
Although the applicability of our results to other groups of somatic symptom disorder is hypothetical and needs to be supported by further studies, this idea is also consistent with recent proposals for the pathophysiology of chronic pain. Here, active inference models of chronic pain have been proposed that largely mirror those that have been proposed for FND (Hechler, Endres, & Thorwart, Reference Hechler, Endres and Thorwart2016; Seymour, Reference Seymour2019). The widely used concept of ‘central sensitisation’ in chronic pain, is entirely compatible with the computational process of abnormal high-level priors relating to pain, which then distort pain perception. Though the word ‘sensitisation’ suggests abnormal sensitivity to incoming sensory/nociceptive input, recent computational models of chronic pain as well as experimental data showing, for example, higher pain thresholds to electrically induced peripheral pain in people with chronic pain, propose a systematic down-weighting of peripheral sensory input and therefore a percept driven by the abnormal high level prior (Hechler et al., Reference Hechler, Endres and Thorwart2016). This is identical to what is proposed in models of FND (Edwards et al., Reference Edwards, Adams, Brown, Parees and Friston2012; Van den Bergh et al., Reference Van den Bergh, Witthoft, Petersen and Brown2017). Similarly, anxiety and depression also fit in the predictive coding model. The role of active inference and predictive coding in emotion processing and depression has already been postulated (Barrett, Quigley, & Hamilton, Reference Barrett, Quigley and Hamilton2016; Lindquist & Barrett, Reference Lindquist and Barrett2012). According to a Dual system fear and anxiety theory, subcortical changes in the brain and body physiology can be modulated by anxiolytics or antidepressants while different cortical networks generating conscious feeling states reflected in self-reports of fear and anxiety can be targeted by psychotherapeutic approaches (LeDoux & Pine, Reference LeDoux and Pine2016).
Clinical implications
What are the clinical implications of the absence of clusters and finding of such a strong intercorrelation of motor and non-motor symptoms severity?
First, it suggests that mechanistic and therapeutic advances in the field of FND, chronic pain and other somatic symptoms may be able to be usefully combined with insights from one symptom type likely to be informative for others.
Second, future revisions of DMS-5 and ICD-11 should consider developing a single diagnostic category covering the full spectrum of ‘functional’ symptoms including pain, fatigue or cognitive complaints. For ICD-11 this should ideally be within both the ‘physical’ and ‘mental’ parts of the classification system, or perhaps more radically within a single ‘brain’ section rather than perpetuating a scientifically and clinically indefensible dualism between brain and mind. This does not imply that neurological and psychiatric illnesses are all best understood at a neurobiological level of understanding, but simply that the brain (and wider nervous system) is the key biological substrate from which neurological, cognitive, emotional and behavioural dysfunction arises.
Third, clinical services might benefit from a degree of unification too. Currently, it is common for services to operate in a rather atomised fashion with chronic pain, chronic fatigue, persistent physical symptoms and FND services working in isolation, alongside multiple speciality-specific services such as functional breathing disorders services in respiratory medicine departments and functional gastrointestinal disorders services within gastroenterology departments. There clearly remains a role for organ-specific specialism in diagnosis and some aspects of treatment. Overlap between functional and organ-specific disease/illness is quite common, meaning that diagnostic expertise within particular medical sub-specialities remains very important (Stone et al., Reference Stone, Carson, Duncan, Roberts, Coleman, Warlow and Sharpe2012). However, there are also many areas of overlap where scientific and clinical skills and knowledge can be pooled. Crucially, rather than considering this as an isolated sub-specialism (such as psychosomatic medicine), such services need to be fully integrated into regular medical practice, which includes the integration of psychiatry and psychology too.
Limitations
Our cluster analysis study should be considered as preliminary, for a more definite conclusion on motor FND subtypes large, multi-centre, international and well-characterised cohorts of patients should be performed. A limitation of this study was the lack of a disease-specific tool for the assessment of subjective motor symptom severity. We used a non-validated simple Likert scale questionnaire tool which may have led to overvaluation of subjective severity in the context of multiple mild symptoms and undervaluation of severely bothersome monosymptomatic manifestations (the more symptoms you are present the higher the score).
Finally, selected measures targeted some of the most common symptoms, however, other important symptoms or aspects of motor FND (e.g. alexithymia, bladder and bowel symptoms etc., dissociative symptoms, sleep disorders) could have been omitted.
Conclusions
This is the first cluster analysis-based study of motor and non-motor symptoms from a relatively large cohort of patients with motor FND. Lack of distinctive subtypes along with a high degree of correlation between all subjective and objective measures of motor, non-motor symptoms and quality of life can be interpreted within the current neurobiological models suggesting unified pathophysiology of the full range of functional symptoms. Our results should inform future revisions of the disease classifications and support the development of a single diagnostic category encompassing patients with FND and other functional somatic symptoms which has important implications for research and service development.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0033291721005225.
Acknowledgements
We thank Irena Starkova for administrative support.
Financial support
This study was supported by project AZV NU20-04-00332.
Conflict of interest
None declared.
