Cognitive training interventions for dementia and mild cognitive impairment in Parkinson’s disease

Vasiliki Orgeta; Kathryn R McDonald; Ellen Poliakoff; John Vincent Hindle; Linda Clare; Iracema Leroi

doi:10.1002/14651858.CD011961.pub2

Intervenciones de entrenamiento cognitivo para la demencia y el deterioro cognitivo leve en la enfermedad de Parkinson

Authors' declarations of interest

Version published: 26 February 2020 Version history

https://doi.org/10.1002/14651858.CD011961.pub2

Collapse all Expand all

Resumen

available in

Antecedentes

Aproximadamente entre el 60% y el 80% de los pacientes con enfermedad de Parkinson (EP) presentan deterioro cognitivo que afecta su calidad de vida. El declive cognitivo es una característica esencial de la enfermedad y a menudo se puede presentar antes de la aparición de los síntomas motores. El entrenamiento cognitivo puede ser una intervención no farmacológica útil que podría ayudar a mantener o mejorar la cognición y la calidad de vida de los pacientes con demencia (DEP) o deterioro cognitivo leve relacionado con la enfermedad de Parkinson (EP‐DCL).

Objetivos

Determinar si el entrenamiento cognitivo (dirigido a uno o varios dominios) mejora la función cognitiva en los pacientes con DEP y EP‐DCL u otras formas claramente definidas de deterioro cognitivo en los pacientes con EP.

Métodos de búsqueda

Se realizaron búsquedas en el registro de ensayos del Grupo Cochrane de Demencia y Trastornos Cognitivos (Cochrane Dementia and Cognitive Improvement Group) (8 de agosto de 2019), en el Registro Cochrane Central de Ensayos Controlados (Cochrane Central Register of Controlled Trials, CENTRAL), en MEDLINE, Embase, CINAHL y PsycINFO. Se realizaron búsquedas en las listas de referencias y en los registros de ensayos, se buscaron revisiones relevantes en el área y en las actas de los congresos. También se estableció contacto con expertos para obtener aclaraciones sobre los datos y los ensayos en curso.

Criterios de selección

Se incluyeron los ensayos controlados aleatorizados en los que los participantes presentaban DEP o EP‐DCL, y en los que la intervención tenía como objetivo entrenar áreas generales o específicas de la función cognitiva, dirigidas a un solo dominio o a múltiples dominios de la cognición, y se comparó con una condición de control. Se consideraron elegibles las intervenciones de componentes múltiples que también incluyeran elementos motores o de otro tipo.

Obtención y análisis de los datos

Dos autores de la revisión, de forma independiente, examinaron los títulos, resúmenes y artículos de texto completo para su inclusión en la revisión. Dos autores de la revisión también realizaron, de manera independiente, la extracción de datos y la evaluación de la calidad metodológica. Para evaluar la calidad general de la evidencia se utilizaron los criterios GRADE.

Resultados principales

Para esta revisión, siete estudios, con un total de 225 participantes, cumplieron los criterios de inclusión. Los siete estudios compararon los efectos de una intervención de entrenamiento cognitivo con una intervención control al final de los períodos de tratamiento que duraron de cuatro a ocho semanas. Seis estudios incluyeron a pacientes con EP que vivían en la comunidad. Estos seis estudios reclutaron a pacientes con deterioro cognitivo leve en un solo dominio (ejecutivo) o en múltiples dominios en la EP. Cuatro de estos estudios identificaron a los participantes con DCL utilizando criterios diagnósticos establecidos, y dos incluyeron a pacientes con EP‐DCL y a pacientes con EP que no presentaban deterioro cognitivo. Un estudio reclutó a pacientes con diagnóstico de demencia en la EP que vivían en instalaciones de cuidados a largo plazo. La intervención de entrenamiento cognitivo en tres estudios se dirigió a un solo dominio cognitivo, mientras que en cuatro estudios se dirigieron a múltiples dominios de la función cognitiva. Los grupos de comparación no recibieron intervenciones o participaron en actividades recreativas (deportes, música, artes), ejercicios de habla o lenguaje, terapia motora computarizada o rehabilitación motora combinada con actividad recreativa.

No se encontró evidencia clara de que el entrenamiento cognitivo mejorara la cognición global. Aunque el entrenamiento cognitivo se asoció con mayores puntuaciones en la cognición global al final del tratamiento, el resultado fue poco preciso y no fue estadísticamente significativo (seis ensayos, 178 participantes, diferencia de medias estandarizada [DME] 0,28, intervalo de confianza [IC] del 95%: ‐0,03 a 0,59; evidencia de certeza baja. No hubo evidencia de una diferencia al final del tratamiento entre el entrenamiento cognitivo y las intervenciones control sobre la función ejecutiva (cinco ensayos, 112 participantes; DME 0,10; IC del 95%: ‐0,28 a 0,48; evidencia de certeza baja) o el procesamiento visual (tres ensayos, 64 participantes; DME 0,30; IC del 95%: ‐0,21 a 0,81; evidencia de certeza baja). La evidencia favoreció al grupo de entrenamiento cognitivo sobre la atención (cinco ensayos, 160 participantes; DME 0,36; IC del 95%: 0,03 a 0,68; evidencia de certeza baja) y la memoria verbal (cinco ensayos, 160 participantes; DME 0,37; IC del 95%: 0,04 a 0,69; evidencia de certeza baja), pero estos efectos fueron menos seguros en los análisis de sensibilidad que excluyeron un estudio en el que sólo una minoría de la muestra presentaba deterioro cognitivo. No hubo evidencia de diferencias entre los grupos de tratamiento y control en las actividades cotidianas (tres ensayos, 67 participantes; DME 0,03; IC del 95%: ‐0,47 a 0,53; evidencia de certeza baja) o la calidad de vida (cinco ensayos, 147 participantes; DME ‐0,01; IC del 95%: ‐0,35 a 0,33; evidencia de certeza baja). Hubo muy poca información sobre los eventos adversos. Se considera que la certeza de la evidencia de todos los resultados fue baja debido al riesgo de sesgo en los estudios incluidos y a la imprecisión de los resultados.

Se identificaron seis ensayos en curso que reclutaron participantes con EP‐DCL, pero no se identificaron ensayos en curso de entrenamiento cognitivo en pacientes con DEP.

Conclusiones de los autores

Esta revisión no encontró evidencia de que los pacientes con EP‐DCL o DEP que reciben entrenamiento cognitivo durante cuatro a ocho semanas presenten mejorías cognitivas importantes al final del entrenamiento. Sin embargo, esta conclusión se basó en un pequeño número de estudios con pocos participantes, limitaciones en el diseño y la realización y resultados poco precisos. Se necesitan estudios más sólidos y con un poder estadístico adecuado sobre el entrenamiento cognitivo antes de poder establecer conclusiones sobre la eficacia del entrenamiento cognitivo en pacientes con DEP y EP‐DCL. Los estudios deben utilizar criterios formales para diagnosticar el deterioro cognitivo, y existe una necesidad particular de más estudios que prueben la eficacia del entrenamiento cognitivo en los pacientes con DEP.

PICOs

Population

Intervention

Comparison

Outcome

The PICO model is widely used and taught in evidence-based health care as a strategy for formulating questions and search strategies and for characterizing clinical studies or meta-analyses. PICO stands for four different potential components of a clinical question: Patient, Population or Problem; Intervention; Comparison; Outcome.

See more on using PICO in the Cochrane Handbook.

Resumen en términos sencillos

available in

Intervenciones de entrenamiento cognitivo para la demencia y el deterioro cognitivo leve en la enfermedad de Parkinson

Pregunta de la revisión

Se deseaba saber si las intervenciones de entrenamiento cognitivo son efectivas para mejorar la cognición (el pensamiento) en los pacientes con demencia o deterioro cognitivo leve en la enfermedad de Parkinson.

Antecedentes

Aproximadamente entre el 60% y el 80% de los pacientes con enfermedad de Parkinson (EP) desarrollan algún grado de deterioro cognitivo, lo que significa que pueden tener dificultades con el pensamiento y el razonamiento, la memoria, el lenguaje o la percepción. Si estas dificultades son lo suficientemente graves como para afectar la capacidad del paciente para realizar las actividades cotidianas, entonces se dice que el paciente presenta demencia de la enfermedad de Parkinson (DEP). Si el paciente presenta problemas cognitivos, pero las actividades cotidianas no están afectadas de manera significativa, entonces se dice que presenta deterioro cognitivo leve de la enfermedad de Parkinson (EP‐DCL). El entrenamiento cognitivo implica la práctica de habilidades cognitivas como la memoria, la atención y el lenguaje a través de tareas específicas. Puede ayudar a los pacientes con DEP o EP‐DCL a mantener mejores habilidades cognitivas.

Lo realizado

Esta revisión examinó si el entrenamiento cognitivo es efectivo para mejorar resultados como las habilidades cognitivas generales ("cognición global"), la memoria, la atención o la capacidad de realizar las actividades cotidianas en pacientes con EP y demencia o DCL. Se buscaron en la literatura médica los estudios de investigación que compararon a los pacientes que recibían una intervención de entrenamiento cognitivo con los que no la recibieron (un "grupo control"). Sólo se incluyeron los estudios en los que la decisión sobre si alguien recibía o no la intervención de entrenamiento cognitivo se tomó de forma aleatorizada; dichos estudios se denominan ensayos clínicos controlados aleatorizados y se consideran el método más justo para comprobar si un tratamiento es eficaz o no. No se examinaron otros tipos de estudios.

Datos encontrados

Se encontraron siete estudios que asignaron al azar un total de 225 participantes a entrenamiento cognitivo o a un grupo control. El tratamiento duró de cuatro a ocho semanas. Todas las intervenciones de entrenamiento cognitivo se realizaron por computadora. Los grupos control no recibieron intervención o recibieron una intervención control, como ejercicios de lenguaje o de motricidad o participación en actividades recreativas. Poco después de terminar el tratamiento no se encontraron diferencias en la cognición global entre los pacientes que recibieron entrenamiento cognitivo y los pacientes de los grupos control. No hubo evidencia convincente de efectos beneficiosos en habilidades cognitivas específicas y no se demostró un efecto beneficioso en las actividades cotidianas o en la calidad de vida. Sin embargo, estas conclusiones se basaron en un número reducido de participantes en un número reducido de estudios. La certeza general de la evidencia fue baja, lo que significa que los resultados de los estudios de investigación posteriores podrían diferir de los resultados de esta revisión.

Conclusión

No se encontró evidencia de buena calidad de que el entrenamiento cognitivo fuera útil en los pacientes con enfermedad de Parkinson y demencia o DCL. Los estudios incluidos fueron pequeños y tenían deficiencias que pueden haber afectado los hallazgos. La certeza de los resultados fue baja y se necesitan más estudios antes de poder confiar en la eficacia del entrenamiento cognitivo en este grupo de pacientes.

Authors' conclusions

Implications for practice

Given that many people with Parkinson's disease (PD) experience cognitive difficulties, the potential benefit of use of cognitive training interventions is large. It will be important that large‐scale trials of effectiveness are conducted, especially in people with PD dementia (PDD). Evidence from this review suggests that cognitive training interventions are generally associated with high levels of adherence. Given the small evidence base, risk of bias, and overall low certainty of the evidence, implications for clinical practice cannot be identified without further research.

Implications for research

Interventions that aim to improve cognition for people with PDD and PD‐related mild cognitive impairment (PD‐MCI) are becoming increasingly important as pharmacological treatment for cognitive symptoms is limited. Our review highlights the paucity of research in the area and that further research is necessary in order to establish whether cognitive training interventions in people with PDD and PD‐MCI may be beneficial. It is important that these trials follow the CONSORT statement. We await evidence from several ongoing trials evaluating the effectiveness of cognitive training in people with PD‐MCI.

Despite some progress in definitions of cognition‐based interventions, there is still confusion with regard to how these interventions are defined and described. Our review suggests overall that although most studies provide enough information to be able to separate interventions, treatment protocols are not provided, and often different approaches are grouped under the same definition. It will be important for future research to concentrate on large‐scale trials of clinical effectiveness versus a control comparison intervention, as many studies had to be excluded because the comparison group was another intervention. Updates of this review could additionally examine effectiveness of cognitive training against other active interventions.

Future studies should define cognitive impairment and specify type of diagnosis. Most studies to date exclude 'people with severe cognitive impairment', so it is likely that in some studies people with mild dementia or MCI were included. It will be important for future research to address this limitation and describe samples in greater detail reporting diagnostic criteria. Future studies should also try to address variation in training period, tasks used (i.e. cognitive only, cognitive and motor tasks), and sensitivity to change of outcome measures.

It will be important to examine whether any effects observed are generalised to everyday function and tasks of daily living. In the current review we are unable to draw any conclusions as to whether cognitive training is associated with improvements in daily life due to the small number of included studies. Future studies should test whether type of intervention evaluated influences efficacy, and comparisons between multicomponent cognitive training interventions, incorporating additional elements such as transfer training and physical training, versus cognitive training alone.

It has been argued that cognitive training may activate mechanisms of cerebral plasticity and slow PD‐associated cognitive decline (Boller 2004). Future studies should investigate the potential mechanisms through which cognitive training may mediate effects on cognition using structural and functional imaging methods, and the extent to which these interventions may slow the progression of cognitive decline (Boller 2004). Research in the area should also address heterogeneity in outcomes, which may hinder future meta‐analyses. In conclusion, our review suggests that there is an urgent need for further large‐scale studies of cognitive training for people with PDD and PD‐MCI.

Summary of findings

Open in table viewer

Summary of findings for the main comparison. Cognitive training compared to control intervention for cognition in PDD and PD‐MCI

Cognitive training compared to control intervention for cognition in PDD and PD‐MCI
Patient or population: cognition in PDD and PD‐MCI Setting: community and long‐term care Intervention: cognitive training Comparison: control intervention (no intervention, participating in recreational activities, receiving speech or language exercises, computerised motor therapy, or motor rehabilitation combined with recreational activity)
Outcomes	SMD (95% CI) meta‐analysis	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
Global cognition post‐treatment Assessed with: MMSE, CERAD Follow‐up: range 4 weeks to 8 weeks	SMD 0.28 higher (0.03 lower to 0.59 higher)	178 (6 RCTs)	⊕⊕⊝⊝ LOW ^{1 2}	A higher score is indicative of improved cognition.
Executive function post‐treatment Assessed with: Trail Making Test B, Stockings of Cambridge Follow‐up: range 4 weeks to 6 weeks	SMD 0.1 higher (0.28 lower to 0.48 higher)	112 (5 RCTs)	⊕⊕⊝⊝ LOW ^{1 2}	A higher score is indicative of improved executive function.
Attention post‐treatment Assessed with: Stroop Task and Brief Test of Attention Follow‐up: range 4 weeks to 6 weeks	SMD 0.36 higher (0.03 higher to 0.68 higher)	160 (5 RCTs)	⊕⊕⊝⊝ LOW ^{1 2}	A higher score is indicative of improved attention.
Verbal memory post‐treatment Assessed with: WMS Logical Memory test, Selective Reminding Test, Hopkins Verbal Learning Test‐Revised, Verbal short‐term memory DemTect Follow‐up: range 4 weeks to 6 weeks	SMD 0.37 higher (0.04 higher to 0.69 higher)	160 (5 RCTs)	⊕⊕⊝⊝ LOW ^{1 2}	A higher score is indicative of improved memory.
Visual processing post‐treatment Assessed with: Judgement Line Orientation Test Follow‐up: range 4 weeks to 6 weeks	SMD 0.3 higher (0.21 lower to 0.81 higher)	64 (3 RCTs)	⊕⊕⊝⊝ LOW ^{1 2}	A higher score is indicative of improved visual processing.
Activities of daily living post‐treatment Assessed with: Barthel Index, Unified Parkinson's Disease Rating Scale, Cognitive Difficulties Scale Follow‐up: range 4 weeks to 8 weeks	SMD 0.03 higher (0.47 lower to 0.53 higher)	67 (3 RCTs)	⊕⊕⊝⊝ LOW ^{1 2}	A higher score is indicative of improved activities of daily living.
Quality of life post‐treatment Assessed with: Parkinson's Disease Questionnaire, QUALIDEM Follow‐up: range 4 weeks to 8 weeks	SMD 0.01 lower (0.35 lower to 0.33 higher)	147 (5 RCTs)	⊕⊕⊝⊝ LOW ^{1 2}	A higher score is indicative of improved quality of life.
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RCT: randomised controlled trial; SMD: standardised mean difference
GRADE Working Group grades of evidence High certainty: We are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: We are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: Our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: We have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.
¹Downgraded one point due to risk of bias (all studies have at least two domains at unclear risk of bias with none of the studies at low risk of bias in all domains). ²Downgraded one point for imprecision due to small sample size (< 400 participants) and wide confidence intervals. PDD: Parkinson's disease dementia PD‐MCI: Parkinson's Disease–Mild Cognitive Impairment MMSE: Mini‐Mental State Examination CERAD: Consortium to Establish a Registry for Alzheimer's Disease test battery WMS: Wechsler Memory Scale

Background

Parkinson's disease (PD) is a common neurodegenerative disorder characterised by motor features such as resting tremor, rigidity, bradykinesia, and postural instability (Hughes 1992). It is now widely accepted that in addition to the motor symptoms, cognitive impairment is a core feature of the disease and should be considered when managing symptoms (Meireles 2012). Cognitive impairment in PD increases in frequency over time, but is already common in the early stages of the disease (Dubois 1997), and may even be present prior to the onset of motor symptoms (Pont‐Sunyer 2015). Longitudinal studies show that people with Parkinson's disease have a three to six times higher risk of developing dementia than the general population without PD (Aarsland 2001). Cognitive impairment negatively affects patient quality of life (Leroi 2012a; Schrag 2000), and increases the risk of developing psychosis (Aarsland 2007; Giladi 2000). Once PD dementia (PDD) develops, the risk of requiring long‐term care increases (Vossius 2011), with substantially higher healthcare costs compared to people living with PD but no dementia (Aarsland 2000).

Current pharmacological treatments include the cholinesterase inhibitors rivastigmine and donepezil. These may reduce some of the direct effects of the disease, including cognitive symptoms (Emre 2004; Ravina 2005; Rolinski 2012), but may be associated with adverse effects (Cutson 1995; van Laar 2011). Memantine is also used. It has a small clinical benefit, is generally safe and well‐tolerated (Leroi 2009; McShane 2019), and may prolong survival (Leroi 2014; Stubendorff 2014). However, clinical responses to drug treatments can vary (Cutson 1995; van Laar 2011). Given the limited number of treatment options and the negative impact of cognitive symptoms in PD, there may be a therapeutic role for non‐pharmacological interventions that target cognitive symptoms. Despite numerous cognition‐based interventions such as cognitive training being investigated in dementia due to Alzheimer's disease (AD) and in mild cognitive impairment (MCI) (Bahar‐Fuchs 2019; Neely 2009; Rojas 2013), less research has focused on the efficacy of cognition‐based interventions in people with cognitive impairment in PD.

Description of the condition

Parkinson's disease is a neurodegenerative disorder affecting approximately 1% to 2% of people aged 60 years and over (de Rijk 2000). Although widely regarded as a motor disorder, it is frequently associated with dementia (Emre 2003; Levy 2002b), which may be a distinct syndrome (Aarsland 2001; Buter 2008). Population‐based studies show that point prevalence of PDD is close to 30%, with dementia incidence rates four to six times higher than in otherwise healthy older people (Emre 2007; Hely 2008). Estimates vary, but the incidence rate of dementia per 1000 persons‐per year amongst people with PD is around 30 (Williams‐Gray 2007), with a cumulative prevalence rate up to 80% (Buter 2008; Hely 2008). The rate of cognitive decline can be represented as an average of a one‐point decrease per year on the Mini‐Mental State Examination (MMSE) (Aarsland 2004). Key risk factors for developing dementia are older age, male sex, more severe stage of parkinsonism, presence of depression, and cognitive symptoms severe enough to meet criteria for MCI (Dubois 2007; Emre 2007; Marinus 2018). PDD is associated with high levels of disability, impaired quality of life, and greater burden of care (Bassett 2005; Leroi 2012a; Vatter 2018). The clinical features of PDD are similar to those in dementia with Lewy bodies (DLB), with both typically involving progressive executive dysfunction, difficulties with visuo‐spatial tasks, and memory impairment Lippa 2007. DLB is diagnosed when cognitive impairment precedes parkinsonian motor signs or is evident within one year from its onset, whereas in PDD, cognitive impairment develops within the context of a well‐established PD diagnosis Emre 2007.

The Movement Disorder Society (MDS) has proposed clinical diagnostic criteria for possible and probable PDD (Emre 2007), providing practical guidance for clinicians and researchers Dubois 2007. These include: (i) a diagnosis of Parkinson's disease according to the Queen Square Brain Bank criteria (Hughes 1992); and (ii) development of motor symptoms prior to dementia onset (McKeith 2002). Dementia is defined as: (a) impairment in at least two cognitive domains, representing a decline from premorbid functioning; and (b) cognitive deficits severe enough to impair daily life, independent of impairment in PD‐related motor symptoms (Dubois 2007). The cognitive profile of PDD is distinct from that of AD, characterised primarily by impairments in attention, executive, and visuo‐spatial functions, but fewer impairments in language compared to AD (Bronnick 2007; Emre 2003). Memory impairment also differs (Aretouli 2010), with greater deficits in retrieval and fewer difficulties with encoding (Aarsland 2003b; Jacobs 1995). A further clinical feature distinguishing PDD from AD is cognitive fluctuations, which are more frequent in PDD (Galvin 2006). Neuropsychiatric symptoms differ between the two populations, with visual hallucinations and sleep disorders occurring more often in PDD (Aarsland 2001).

Lewy body‐type degeneration is considered to be the main pathology (Halliday 2014; Irwin 2012); however, cortical changes typical of AD and frontal atrophy may also be present (Lashley 2008; Sabbagh 2009). Neurochemically, cholinergic deficits are also found in PDD as well as in AD. This provides the rationale for the use of cholinesterase inhibitors in PDD (Aarsland 2003a; Emre 2004; Leroi 2004; Reading 2001), which may improve cognition and activities of daily living (Rolinski 2012).

Cognitive impairment severe enough to meet criteria for mild cognitive impairment is frequent in Parkinson's disease (Aarsland 2010), and increases the risk of developing PDD (Pedersen 2013). The term 'Parkinson's disease with mild cognitive impairment' (PD‐MCI) has been proposed by the Movement Disorder Society Task Force alongside consensus‐derived clinical diagnostic criteria that remain to be validated (Litvan 2012). PD‐MCI criteria include the following:

subjective report of cognitive problems by the patient or carer;
performance at least 1.5 standard deviations (SDs) below the age‐corrected mean score in one cognitive domain;
no impairments in activities of daily living that can be attributed to cognitive impairment.

Approximately 50% of people with PD have mild cognitive impairment (Aarsland 2010; Janvin 2005), with more than 40% presenting with PD‐MCI at time of diagnosis (Aarsland 2010; Yarnall 2014). PD‐MCI primarily affects memory, visuo‐spatial, and executive functions and may be a transitional state between normal ageing and dementia (Backman 2005). There may also be a subtype of PD‐MCI that is non‐progressive and does not convert to dementia (Williams‐Gray 2007), although the majority of individuals with PD‐MCI progress to PDD over time (Caviness 2007; Janvin 2006; Williams‐Gray 2009). As with PDD, PD‐MCI is associated with older age at disease onset, being male, experiencing depression, and having severe motor symptoms (Aarsland 2010).

Description of the intervention

For the purposes of this review, we systematically reviewed cognitive training interventions in people with PDD and PD‐MCI. Cognition‐based interventions differ from other non‐pharmacological interventions in that they specifically aim to enhance cognition, as opposed to other behavioural or functional outcomes, either directly or indirectly. Given that terminology describing these interventions can be confusing, in this review we followed the classification of Bahar‐Fuchs 2019, dividing these interventions into three types:

cognitive stimulation;
cognitive training;
cognitive rehabilitation.

Interventions that involve non‐specific stimulation of cognitive and social functioning, engaging patients in general activities and discussions are termed 'cognitive stimulation' approaches, whereas 'cognitive training' interventions use guided practice on standardised paper and pencil or computerised tasks to target specific areas of cognition (single or multiple domains). 'Cognitive rehabilitation' uses individualised approaches to target restrictions in everyday life and improve functioning in relation to individualised goals. Interventions may also use mixed approaches, combining elements across cognition‐based approaches or by adding additional physical or motor components. In this review we included studies of cognitive training targeting either a single cognitive domain or multiple cognitive domains.

Depending on the protocol used, cognitive training targets a single or multiple cognitive domains, for example memory, executive, attention, and visuo‐spatial functions, and is delivered face‐to‐face or remotely, with sessions lasting from 30 minutes to an hour. Tasks may vary in complexity, and may be individually tailored (taking into account baseline cognitive performance or level of cognitive impairment) (Calleo 2012). Standardised programmes have been developed, as well as multimodal interventions, incorporating training in everyday activities and practising daily tasks using mnemonics, planning, and memory training. Interventions may take place in various settings (outpatient clinics, hospital settings, person's own home), on an individual or group basis, using either paper and pencil or multimedia computer software.

How the intervention might work

Cognition‐based interventions in people with cognitive impairment of any aetiology (e.g. neurodegenerative disorders, traumatic brain injury, stroke), have been guided theoretically by restorative or compensatory approaches; these involve respectively either improving specific cognitive functions or using contextualised perspectives in which training is adapted to accommodate cognitive impairment (Ylvisaker 2002). For example, people engage in specific tasks that target one or more areas of cognitive function through guided practice (Bahar‐Fuchs 2019). These types of interventions have been associated with improvements in memory in healthy older people and people with non‐PD‐related mild cognitive impairment, although they have not been shown to outperform active control interventions (Lampit 2014; Martin 2011). Generalisation of effects to other outcomes such as activities of daily living is still limited in both cognitively healthy older people and in people with mild cognitive impairment (Kelly 2014; Reijnders 2007).

In line with restorative approaches, cognitive training strengthens neural networks of attentional and control processes via neuroplasticity, as a result of experience or environmental stimulation (Raz 2006; Shaw 1994). Cognitive training may enhance frontal lobe function by activating mechanisms of brain plasticity (Boller 2004). Animal and human studies have shown that sensory systems in the cerebral cortex can improve through learning and practising tasks and that brain changes in cortical areas mediate cognitive improvements (Buonomano 1998; Gilbert 2001). Training in specific tasks increases grey matter volume (Driemeyer 2008). Cognitive training exercises increase memory‐related activation in several brain areas in people with non‐PD‐related mild cognitive impairment such as memory‐related hippocampal function (Belleville 2011; Hampstead 2012), consistent with the notion that cognitive training may encourage neuroplasticity of the brain.

Why it is important to do this review

Whilst treatment for motor symptoms in PD has improved considerably, treatment of cognitive symptoms remains limited. In clinical practice, PDD is often under‐recognised and not optimally managed. The effects of drug treatments on symptoms are modest (Aarsland 2009; Emre 2010; Horstink 2006; Leroi 2009; Rolinski 2012), and no disease‐modifying therapy is available. Polypharmacy, high medical comorbidity, and the side effect profile of the drugs all contribute to problems with tolerability of cholinesterase inhibitors, limiting access to evidence‐based treatments for some people with PDD and PD‐MCI (Rolinski 2012).

Cognitive impairment in the context of Parkinson's disease increases morbidity and mortality and experiences of neuropsychiatric symptoms (Buter 2008; Hughes 1992; Leroi 2012b), and is a marker of rapid functional decline (Hely 1995). Cognitive decline decreases patient quality of life (Leroi 2012a; Levy 2002a), and increases carer burden (Leroi 2012b), therefore any interventions that alleviate cognitive symptoms have the potential to reduce disability and improve quality of life for people with PDD and their carers.

Although cognitive training may be useful in improving cognitive outcomes in PDD, its effectiveness has not been systematically reviewed. It is important to describe the effects of these interventions in this population separately from others with dementia, as cognitive deficits are different to those observed in other dementias, and therefore interventions may need to be tailored to the cognitive domains commonly affected in PD. Since many people with PD present with MCI at the time of diagnosis (Smith 1999), identifying cognitive training approaches that can help manage cognitive symptoms in PD‐MCI would be very useful clinically. The current review aims to benefit clinical practice by identifying whether cognitive training interventions improve cognitive function in people with PDD and PD‐MCI, and to make recommendations for future research.

Objectives

Primary objective

To determine whether cognitive training (targeting single or multiple domains) improves cognition in people with PDD and PD‐MCI or other clearly defined forms of cognitive impairment in people with PD.

Secondary objectives

To determine the effect of cognitive training on quality of life, activities of daily living, neuropsychiatric symptoms, adverse events, carer quality of life, and carer burden.

Methods

Criteria for considering studies for this review

Types of studies

We included studies that fulfilled the following criteria:

were randomised controlled trials (RCTs), including cluster‐randomised trials;
included a control or comparison group receiving no specific cognitive intervention.

Types of participants

People of any age, from any setting (e.g. home, community, long‐term care, or rehabilitation settings), diagnosed with PDD (Emre 2007), or people with PD‐MCI (Litvan 2012). We included studies that used the criteria for MCI proposed by Petersen 1999 or similar criteria.

Types of interventions

We included studies that reported a comparison between a cognitive training intervention and a control intervention. Cognitive training was defined as any intervention that targeted cognition (single or multiple cognitive domains) using a cognitive training approach involving guided practice (Bahar‐Fuchs 2019; Davis 2001). Cognitive training interventions could be of any intensity, duration, or frequency, conducted on an individual or group basis, with or without the involvement of carers. We did not exclude trials on the basis of the language used to describe the intervention in the trial paper. Cognitive training interventions meeting criteria for inclusion in this review could also be described as 'memory therapy' or 'cognitive therapy', 'cognitive groups', 'cognitive training or retraining', 'cognitive support', or 'cognitive stimulation'.

Eligible control conditions could include no treatment (usual care), a waiting list for cognitive training, or an active control condition in which the comparison group engaged in non‐specific activity (i.e. an attention control, controlling for effects of staff attention or social contact). 'Usual care' refers to what would usually be provided to people with cognitive impairment in Parkinson's disease in the setting in which the study was conducted (including medication, day care, and support, but no specific structured cognitive training intervention). Multicomponent interventions were considered eligible as long as one component was a clearly defined cognitive training intervention. We imposed no restrictions regarding additional physical or motor components. We excluded treatments identified as exercise, music, art, befriending, or bibliotherapy.

Types of outcome measures

We included studies that reported a cognitive outcome or outcomes, measured by a standardised test or any test that has acceptable psychometric properties.

Primary outcomes

Measures of cognitive function: global cognition, executive function, attention, memory (specifically verbal memory), and visual processing.

Secondary outcomes

Measures of function (e.g. activities of daily living)
Measures of quality of life
Measures of neuropsychiatric symptoms including depression, anxiety, and apathy assessed by a validated rating scale
Measures of carer outcomes including quality of life, experience of carer burden, well‐being, or mood
Adverse effects (e.g. on mood, awareness of cognitive difficulties)

We included studies assessing outcomes during or immediately after the intervention period.

Search methods for identification of studies

Electronic searches

We searched ALOIS (alois.medsci.ox.ac.uk/), the Cochrane Dementia and Cognitive Improvement Group’s Specialised Register, on 8 August 2019. ALOIS is maintained by the Information Specialists of the Cochrane Dementia and Cognitive Improvement Group and contains studies in the areas of dementia (prevention and treatment), mild cognitive impairment, and cognitive improvement. The studies are identified from:

monthly searches of a number of major healthcare databases: MEDLINE, Embase, CINAHL (Cumulative Index to Nursing and Allied Health Literature), PsycINFO, and LILACS (Latin American and Caribbean Health Science Information database);
monthly searches of a number of trial registers: ISRCTN, UMIN (Japan's trial register), and the World Health Organization International Clinical Trials Registry Platform (WHO ICTRP) (which covers the US National Institutes of Health Ongoing Trials Register ClinicalTrials.gov, ISRCTN, the Chinese Clinical Trials Register, the German Clinical Trials Register, the Iranian Registry of Clinical Trials, and the Netherlands National Trials Register, plus others);
quarterly search of the Cochrane Central Register of Controlled Trials (CENTRAL) in the Cochrane Library;
six‐monthly searches of a number of grey literature sources from ISI Web of Science Core Collection.

To view a list of all sources searched for ALOIS see 'About ALOIS' on the ALOIS website. Details of the search strategies used for the retrieval of reports of trials from the healthcare databases, CENTRAL, and conference proceedings can be viewed in the ‘Methods used in reviews’ section within the editorial information about the Dementia and Cognitive Improvement Group. We performed additional searches in many of the sources listed above to cover the time frame from the last searches performed for ALOIS to ensure that the search for the review was as up‐to‐date and comprehensive as possible. We also searched relevant reviews in the area and conference proceedings.

The search strategies used are described in Appendix 1. We have run five separate searches to date, the latest one on 8 August 2019.

Searching other resources

We contacted corresponding authors of identified ongoing trials for additional references and unpublished data. We scanned the reference lists of identified publications for additional trials, and all review papers related to cognitive training interventions in cognitive impairment in PD, PDD, and PD‐MCI.

Data collection and analysis

Selection of studies

Two review authors (VO, KM) independently assessed the titles and abstracts of records identified by the search for potential eligibility. Any disagreements were resolved by discussion with a third review author (IL). We excluded studies that did not meet the inclusion criteria and obtained full‐text copies of those references deemed potentially relevant. We documented reasons for the exclusion of studies. Where necessary, we requested additional information from the original study authors, specifically for ongoing trials and studies reporting mixed data on people with PD with or without cognitive impairment. We repeatedly contacted authors to request further information and were successful in one instance.

Data extraction and management

We extracted information about methods, participants, interventions, outcomes, and results as described below for all studies meeting the inclusion criteria, ongoing studies, and studies awaiting classification.

Participants: Characteristics of the sample (age, diagnostic criteria, severity of cognitive impairment, and exclusion criteria).
Methods: Data were extracted on methodologies used for randomisation, blinding, and participant dropout.
Interventions: Duration, intensity, type, and frequency of cognitive training and control interventions.
Outcomes: Primary outcomes included measures of cognition. Secondary outcomes were activities of daily living, quality of life, and adverse events. Secondary outcome measures for carers were quality of life, experience of carer burden, carer well‐being or mood.
Results: Where data were available, we collected the number of participants on whom the outcome was measured in each group, means and SDs. We used change from baseline scores for all of the analyses reported and calculated the change scores manually. Calculations of the SD of change scores were based on an assumption that the correlation between measurements at baseline and those at subsequent time points is zero. This method overestimates the SD of the change from baseline, but is considered preferable in a meta‐analysis to take a conservative approach. For one study (three analyses) (Folkerts 2018), we used median scores of outcomes and calculated the SD from the interquartile range.

Assessment of risk of bias in included studies

We used Cochrane's tool for assessing risk of bias to evaluate the methodological quality of the included studies (Higgins 2011). The tool addresses six specific domains: sequence generation, allocation concealment, blinding, incomplete outcome data, selective reporting, and other sources of bias. Two review authors (VO, KM) independently assessed each domain, resolving any differences by discussion with a third review author (IL). In cases where no information was available to make a judgement, this is explicitly stated.

Measures of treatment effect

All outcomes were continuous, and a variety of different scales contributed data to each meta‐analysis, therefore we used the standardised mean difference (SMD) as the measure of treatment effect.

Unit of analysis issues

Where trials had multiple treatment groups, we combined all relevant experimental groups into a single group and all relevant control groups into a single control group. We did not identify any cluster‐RCTs.

Dealing with missing data

We reported the number of participants included in the final analysis as a proportion of all participants in the study.

Assessment of heterogeneity

We used the I² statistic to assess heterogeneity amongst studies. We defined substantial heterogeneity as an I² of more than 50%. We observed no or minimal heterogeneity in all of our analyses (I² ≤ 1%).

Assessment of reporting biases

We only identified seven RCTs and pooled data from a maximum of six studies in total, therefore we did not use a funnel plot to assess for publication bias (Egger 1997).

Data synthesis

Had data permitted, we would had performed separate analyses on PDD and PD‐MCI and analysed cognitive training interventions separately from multicomponent interventions. We did conduct analyses by cognitive domain: specifically, we separately analysed effects of cognitive training on global cognition, executive function, attention, verbal memory, and visual processing ability.

Subgroup analysis and investigation of heterogeneity

We had planned to conduct subgroup analyses comparing the effects of cognitive training versus active or passive control conditions, and comparing the effects of single‐domain versus multiple‐domain cognitive training interventions. However, there were too few studies and participants to permit any meaningful subgroup analyses..

Sensitivity analysis

We conducted sensitivity analyses excluding studies where fewer than 50% of participants had PD dementia, mild cognitive impairment, or other form of clinically significant cognitive decline as verified by a neuropsychological test.

Summary of findings and assessment of the certainty of the evidence

We used the GRADE approach to assess the certainty of the evidence for the included studies reporting on the treatment effect of cognitive training in PD‐MCI and PDD compared to a control condition for a total of seven outcomes (Guyatt 2011). We used risk of bias, imprecision, inconsistency, indirectness, and publication bias to rate the overall certainty of the evidence. We have presented key findings of the review in the summary of findings Table for the main comparison, which includes ratings of the certainty of evidence for all outcomes.

Results

Description of studies

Results of the search

We identified a total of 3849 results via the electronic searching and five articles via other sources. After a first assessment of the search results performed by the Cochrane Dementia and Cognitive Improvement Group Information Specialists, 456 results remained for evaluation (first search: 126 results, four studies identified via handsearch; second search: 21 results, one study identified by handsearch; third search: 54 results; fourth search: 165 results; fifth search: 85 results). We screened a total of 63 full‐text articles for eligibility, of which 47 were excluded with reasons; seven studies met the inclusion criteria (Alloni 2018; Cerasa 2014; Costa 2014; Folkerts 2018; Lawrence 2018; París 2011; Petrelli 2014); six studies are ongoing (NCT03582670; ACTRN12618000999235; NCT02225314; NCT03285347; NCT02525367; van de Weijer 2016); and three studies are awaiting classification until further information is obtained (NCT01647698; NCT01646333; NCT02920632); (see Figure 1).

Figure 1

Study flow diagram.

Included studies

See Characteristics of included studies.

Of the seven studies meeting the inclusion criteria (Alloni 2018; Cerasa 2014; Costa 2014; Folkerts 2018; Lawrence 2018; París 2011; Petrelli 2014), all studies except Costa 2014 contributed to the analysis of effects of cognitive training on global cognition. Five studies contributed to the analysis of effects of cognitive training on executive function (Alloni 2018; Cerasa 2014; Costa 2014; Lawrence 2018; París 2011). Five studies contributed to the analysis of effects on attention and verbal memory (Alloni 2018; Cerasa 2014; Lawrence 2018; París 2011; Petrelli 2014). Three studies contributed to analyses of visual processing ability (Cerasa 2014; Lawrence 2018; París 2011). Three studies contributed to the analysis of activities of daily living (Folkerts 2018; Lawrence 2018; París 2011), and five studies contributed to the analyses of effects of cognitive training on quality of life (Cerasa 2014; Folkerts 2018; Lawrence 2018; París 2011; Petrelli 2014).

Design

All seven studies were RCTs that evaluated the effects of a cognitive training intervention aimed at training one or more domains of cognitive function.

Setting

Alloni 2018, Cerasa 2014, and Costa 2014 were conducted in Italy and recruited outpatients in neurorehabilitation or academic units. Folkerts 2018 recruited residents of a PDD‐specific long‐term care unit in the Netherlands. The study by Lawrence 2018 was conducted in Western Australia and recruited outpatients. París 2011 was conducted in Spain and recruited outpatients from movement disorders clinics. Petrelli 2014 was conducted in Germany and recruited participants from a university hospital, an outpatient movement disorders clinic, and regional PD support groups.

Participants

Thirty‐one participants were randomised in Alloni 2018; 20 in Cerasa 2014; 17 in Costa 2014; 12 in Folkerts 2018; 42 in Lawrence 2018; 33 in París 2011; and 70 in Petrelli 2014.

In the study by Alloni 2018, participants had a diagnosis of idiopathic PD (United Kingdom Parkinson's Disease Society Brain Bank (UKPDBB) criteria) (Hughes 1992), and scored ≤ 4 on the Hoehn and Yahr scale (Hoehn 1967). All participants experienced single‐domain (executive) or multiple‐domain mild cognitive impairment with executive involvement (Litvan 2012). People with pre‐existing cognitive impairment (e.g. aphasia, neglect), severe disturbances in consciousness, psychiatric or neurological conditions, or severe motor or sensory disorders were excluded. The mean Mini‐Mental State Exam (MMSE) score of participants at baseline was 25.35 (SD = 2.59).

In Cerasa 2014, participants had a clinical diagnosis of PD (UKPDBB criteria) and predominant deficits in either attention and/or information processing speed, working memory and/or executive functioning (demonstrated in at least one of the following tests: Symbol Digit Modalities Test (SDMT), Trail Making Test (TMT A–B), Paced Auditory Serial Addition Test (PASAT), digit span forward and backward, and Stroop word‐colour task (ST), but no additional impairment in other cognitive domains (i.e. language, verbal and spatial long‐term memory) or motor complications (i.e. levodopa‐induced dyskinesias) (Hughes 1992). People with dementia (Diagnostic and Statistical Manual of Mental Disorders, fourth edition (DSM‐IV) criteria), psychiatric history (assessed by structured clinical interview of the DSM‐IV), vascular brain lesions, brain tumour and/or marked brain atrophy were excluded. The mean MMSE score of participants at baseline was 29.05 (SD = 1.00).

In Costa 2014, idiopathic PD was defined according to UKPDBB criteria (treated with levodopa or dopamine agonists, or both) (Hughes 1992); all participants had MCI according to Litvan 2012 criteria and performed 1.5 SD below the normative population in two tests of a neuropsychological screening battery (one of which investigated executive functioning). People with psychiatric disorders, neurological conditions other than PD, vascular brain lesions, metabolic disease, or people with significant changes in routine activities (measured by standardised tests) were excluded. The mean MMSE score of participants at baseline was 28.25 (SD = 1.60).

In the study by Folkerts 2018, all participants had PDD according to the MDS criteria (Emre 2007). Inclusion criteria were being a resident in the unit, having idiopathic PD (diagnosed by a neurologist/psychiatrist), experiencing cognitive dysfunction (MMSE score between 10 to 25), good language, vision, and hearing, and consent from a legal representative. People with a history of alcohol or drug abuse, a life‐threatening illness, psychotic symptoms, or those who were bedridden were excluded. The mean MMSE score of participants at baseline was 17.83 (SD = 5.55).

In Lawrence 2018, participants were diagnosed with idiopathic PD (UKPDBB criteria) and MCI in accordance with the MDS PD‐MCI level II diagnostic criteria (Hughes 1992; Litvan 2012), were on a stable response to antiparkinsonian medication, and cognitive deficits did not interfere with functional independence (Unified Parkinson's Disease Rating Scale (UPDRS‐II) score less than 3). People with PDD, a recent history of brain surgery, migraine, or epilepsy were excluded. The mean MMSE score of participants at baseline was 25.66 (SD = 2.08).

In the study by París 2011, participants were diagnosed with PD (UKPDBB criteria) (Hughes 1992), with disease severity of Hoehn and Yahr stages I to III and not receiving any other cognitive, psychological, or physical treatment (Hoehn 1967). People with significant cognitive impairment (MMSE < 23), below‐average premorbid intelligence (vocabulary subtest, Wechsler Adult Intelligence Scale‐III (WAIS‐III) typical score < 40), on cholinesterase inhibitors, or having any change in their medication were excluded. People with major depression (Geriatric Depression Scale (GDS‐15) > 10), severe sensory deficits, or a psychiatric/neurological condition were excluded. Fifty per cent (14 of 28 participants) met Petersen 2005 and Artero 2006 criteria for MCI and demonstrated a decrement of more than 1.5 SDs on a cognitive test or subtest. The mean MMSE score of participants at baseline was 27.85 (SD = 1.37).

In Petrelli 2014, participants had idiopathic PD (UKPDBB criteria) (Hughes 1992). People with suspected dementia (MMSE < 25), other neurological or psychiatric diseases, or impaired hearing or sight were excluded. Twenty per cent of the sample fulfilled criteria for PD‐MCI (Litvan 2011). The mean MMSE score of participants at baseline was 27.90 (SD = 1.93).

Cognitive training interventions and control comparisons

Alloni 2018 evaluated computerised cognitive training (CoRe Alloni 2015) consisting of patient‐tailored exercises aimed primarily at stimulating executive function versus a control intervention incorporating motor rehabilitation combined with recreational activity. The cognitive training was individual sessions of 45 minutes 3 times a week over 4 weeks (12 sessions in total).

Cerasa 2014 investigated the effectiveness of individually tailored group computer‐based attention‐training (RehaCom Cerasa 2013) versus computerised motor therapy consisting of simple visuomotor co‐ordination tasks. Cognitive training was 1‐hour group sessions twice a week for 6 weeks (12 sessions in total).

Costa 2014 compared a cognitive training intervention aimed at practising shifting ability (prospective memory) versus simple language and respiratory exercises (Macdonald 2011). Cognitive training was delivered in 45‐minute sessions, 3 times per week for 4 weeks (12 sessions in total).

Folkerts 2018 evaluated a modified version of the NEUROvitalis senseful programme, which targeted executive and visual spatial function versus treatment as usual, which incorporated a variety of non‐pharmacological interventions such as sports, music, and arts open to all residents (Baller 2009; Middelstadt 2016). The intervention was delivered for 60 minutes twice weekly for 8 weeks (16 sessions in total).

Lawrence 2018 evaluated an interactive online computer‐based cognitive training known as Smartbrain Pro (www.smartbrain.net) (Tárraga 2006), which aims to train several cognitive domains (attention, working memory, psychomotor speed, executive function, and visuo‐spatial ability). The training was delivered in sessions of 45 minutes each, 3 times per week for 4 weeks (12 sessions in total). This was an RCT with six parallel intervention arms: standard cognitive training, tailored cognitive training, transcranial direct current stimulation (tDCS), standard cognitive training + tDCS, or tailored cognitive training + tDCS. We combined the standard cognitive training and the tailored cognitive training intervention groups and compared them to the no‐intervention control group in the meta‐analysis.

París 2011 evaluated the same Smartbrain Pro training, also in 45‐minute sessions, 3 times per week for 4 weeks (12 sessions in total). In París 2011, participants completed additional homework exercises stimulating specific and non‐specific cognitive areas. The comparator was group speech therapy.

Petrelli 2014 compared structured training using the same NEUROvitalis programme as Folkerts 2018, unstructured cognitive training using a programme called "Mentally fit", and a waiting‐list control group. The interventions were delivered as group sessions of 90 minutes, twice a week for 6 weeks (12 sessions in total). We combined the structured and unstructured cognitive training groups and compared them with the waiting‐list control group in the meta‐analysis.

For further details see Characteristics of included studies.

Adherence

Folkerts 2018 reported the highest level of adherence (92.7%) across all studies, followed by Alloni 2018 (90%) and Cerasa 2014 (80%). In both París 2011 and Petrelli 2014, at least 88% of the sample completed over 75% of sessions. Costa 2014 and Lawrence 2018 did not provide details regarding adherence to the intervention.

Outcomes

All studies reported outcomes immediately after the intervention was finished, with two studies also reporting later follow‐up: Alloni 2018 at 24 weeks and Lawrence 2018 at 12 weeks. End‐of‐treatment time points were: four weeks for Alloni 2018, Costa 2014, Lawrence 2018, and París 2011; six weeks for Cerasa 2014 and Petrelli 2014; and eight weeks for Folkerts 2018.

We classified cognitive measures used in each of the studies by considering which instrument contributed most to each outcome in line with the primary outcomes set for the review and the similarity of instruments used across studies. This task involved judgement, as many of the measures in the included studies can relate to several cognitive domains. We extracted change from baseline values for all analyses and outcomes due to imbalances at baseline and small sample sizes across all studies. Details of which outcome measures contributed to each of the analyses appear below.

Primary outcome ‐ Cognition

Global cognition

Global cognition was measured by the MMSE, Folstein 1975, in five studies (Alloni 2018; Cerasa 2014; Lawrence 2018; París 2011; Petrelli 2014); scores on the MMSE range from 0 to 30, with lower scores indicative of greater impairment in cognition. Folkerts 2018 used the Consortium to Establish a Registry for Alzheimer's Disease (CERAD) (Welsh 1994), where higher scores indicate better performance. Three studies, Alloni 2018; París 2011; Petrelli 2014, additionally used the Montreal Cognitive Assessment (MoCA) (Nasreddine 2005), Addenbrooke's Cognitive Examination (ACE) (Mathuranath 2000), and the DemTect (Kalbe 2004), respectively, to measure global cognition. Lower scores indicate greater impairment in cognition in all instruments. Lawrence 2018 also measured cognition using the Parkinson's disease‐cognitive rating scale (Pagonabarraga 2008); higher scores indicate better performance in this scale.

In our meta‐analysis of effects of cognitive training on global cognition, we used the MMSE where possible to reduce heterogeneity and make results comparable to the wider literature on cognitive‐based interventions and pharmacological trials. We used an alternative only if a study had not used the MMSE. For our meta‐analysis on global cognition we included the following measures.

MMSE for Alloni 2018, Cerasa 2014, and Lawrence 2018 (combining the two treatment arms of standard and tailored cognitive training); París 2011 and Petrelli 2014 (combining the two treatment arms of structured and unstructured cognitive training).
CERAD for Folkerts 2018.

Cognitive subdomains

Alloni 2018 measured executive function using the Raven's Matrices 47 test (RM47) (Raven 1988); the Weigl's Colour‐Form Sorting Test (WCFT) (Weigl 1941); the Frontal Assessment Battery (FAB) (Dubois 2000); and the F‐A‐S Test (Spreen 1977). Attention was measured using the Attentive Matrices (Spinnler 1987); Trail Making Tests A and B (Reitan 1985); and the Stroop task (Stroop 1935). Verbal memory was measured with the verbal and digit span tests (Wechsler 2001); Rey's 15‐words test (Rey‐15) (Lezak 1983); and the Wechsler Memory Scale (WMS) Logical Memory test measuring immediate and delayed recall (Wechsler 1945). Spatial memory was assessed using the Corsi Block‐Tapping Test and the Rey‐Osterrieth Complex Figure Test (Corsi 1972; Rey 1941). Visuo‐spatial ability was measured using the Rey‐Osterrieth Complex Figure Copy test (Meyers 1995).

Cerasa 2014 measured executive function using the Controlled Oral Word Association Test (COWAT) (Benton 1983). Attention and working memory were measured by the Symbol Digit Modalities Test (SDMT) (Smith 2007), the Paced Auditory Serial Addition Test (PASAT) (Gronwall 1974), the digit span test (Wechsler 2001), the Trail Making Test A and B and B‐A (Reitan 1985); and the Stroop task (Stroop 1935). Verbal memory was assessed using the Selective Reminding Test (Buschke 1973), spatial memory with the Rey‐Osterrieth Complex Figure Test (Rey 1941), and visuo‐spatial ability with the Judgment of Line Orientation (JLO) test (Benton 1994).

Costa 2014 measured executive function using tasks of phonemic, semantic and alternating (phonemic/semantic) fluency and attention with the Trail Making Test A and B (Downes 1993; Reitan 1985).

Lawrence 2018 assessed executive function with the Stockings of Cambridge (SOC) test, Robbins 1998, and the Controlled Oral Word Association Test (COWAT), Benton 1983. Attention was measured by the Letter Number Sequencing test, Wechsler 1945, and the Stroop task, Stroop 1935. Verbal memory was assessed by the Hopkins Verbal Learning Test‐Revised, Brandt 2001, and the Paragraph Recall Test (PRT), Wilson 1989. Language was assessed with the Boston Naming Test (BNT), Kaplan 1983, and the Similarities Test, Wechsler 1945. Visuo‐spatial ability was measured by the JLO test, Benton 1994, and the Hooper Visual Organization Test (HVOT), Hooper 1983.

París 2011 measured executive function using the Tower of London (TOL), Culberston 2001, and the F‐A‐S Test, Spreen 1977. Attention and working memory were measured with the Wechsler Adult Intelligence Scale (WAIS‐III) Digit Span test (Wechsler 2001), the SDMT (Smith 2007), Trail Making Tests A and B (Reitan 1985), and the Stroop task (Stroop 1935). Verbal memory was assessed by the California Verbal Learning Test Revised, Delis 2000, and the WMS Logical Memory test, Wechsler 1945. Spatial memory was assessed with the Rey‐Osterrieth Complex Figure Copy test, Meyers 1995, and visuo‐spatial ability with the JLO test, Benton 1994.

Petrelli 2014 measured executive function with the digit span reverse test of the DemTect and verbal and semantic fluency with fluency tasks of the same instrument (Kalbe 2004). The Brief test of attention was used to measure attention (Schretlen 1996). Verbal short‐ and long‐term memory and visual long‐term memory were assessed by the DemTect (Kalbe 2004). Visuo‐spatial ability was measured with the Rey‐Osterrieth Complex Figure Copy test (Meyers 1995).

We thoroughly reviewed all tests to ensure there was as much overlap as possible across domains and instruments used.

For analyses addressing effects of cognitive training on executive function, we selected the following tests: Trail Making Test B for Alloni 2018, Cerasa 2014, Costa 2014, and París 2011; and Stockings of Cambridge (SOC) for Lawrence 2018 (combining the two treatment groups).

For analyses of effects on attention, we used the following tests: the Stroop task for Alloni 2018, Cerasa 2014, and Lawrence 2018 (combining the two treatment groups) and París 2011; and the Brief test of attention for Petrelli 2014 (combining the two treatment groups).

For analyses of effects on verbal memory, we used the following tests: the WMS Logical Memory test (Immediate Recall) for Alloni 2018 and París 2011; the Selective Reminding Test‐long‐term storage for Cerasa 2014; the Hopkins Verbal Learning Test‐Revised Immediate Recall subtest for Lawrence 2018 (combining the two treatment groups); and the verbal memory test of short‐term memory of the DemTect for Petrelli 2014 (combining the two treatment groups).

For analyses of effects of cognitive training on visual processing, we included studies that used the JLO test, which were Cerasa 2014, Lawrence 2018 (combining the two treatment groups), and París 2011.

Secondary outcomes

Activities of daily living

Folkerts 2018 measured ADLs using the Barthel Index (Barthel 1965), where higher scores are indicative of greater independence in activities of daily living. Lawrence 2018 used the UPDRS Part II (Goetz 2008), in which higher scores are indicative of more severe impairment. París 2011 evaluated cognitive difficulties in ADLs using the Cognitive Difficulties Scale (CDS) (McNair 1983), where higher total summed scores indicate worse cognitive complaints associated with ADLs. We combined these three studies for our analyses on effects of cognitive training on ADLs (Folkerts 2018; Lawrence 2018 (combining the two treatment groups); París 2011).

Quality of life

Four studies, Cerasa 2014; Lawrence 2018; París 2011; Petrelli 2014, used the Parkinson's Disease Questionnaire (PDQ‐39) to measure quality of life (Jenkinson 1997); lower scores in this scale indicate higher quality of life. Folkerts 2018 used the QUALIDEM scale (Ettema 2007), in which higher scores are indicative of higher quality of life. We pooled data from all five of these studies to investigate the effects of cognitive training on quality of life (Cerasa 2014; Folkerts 2018; Lawrence 2018 (combining the two treatment groups); París 2011; Petrelli 2014 (combining the two treatment groups)).

Depression

Cerasa 2014 and Petrelli 2014 used the Beck Depression Inventory II to measure depressive symptoms (Beck 1996); París 2011 used the Geriatric Depression Scale (GDS) (Sheikh 1986); Folkerts 2018 used the GDS and the Cornell Scale for Depression in Dementia (CSDD) Alexopoulos 1988. Higher scores indicate more symptoms in all measures.

Other outcomes

Cerasa 2014 measured anxiety symptoms with the State‐Trait Anxiety Inventory (STAI) Spielberger 1983. Folkerts 2018 measured health‐related quality of life with the EQ‐5D‐5L EuroQoL Group 1990, and neuropsychiatric symptoms with the Neuropsychiatric Inventory (NPI) Cummings 1997. None of the studies included outcome measures for carers.

For further details see Characteristics of included studies.

Ongoing trials

We identified six ongoing trials, which on the basis of the information available meet the inclusion criteria for this review. These are described in the Characteristics of ongoing studies table.

Studies awaiting classification

We found three studies that are awaiting classification. These are all ongoing trials, but it is unclear whether or not they will meet our review inclusion criteria. For two trials, we await further information regarding the inclusion criteria of participants (whether participants meet criteria for PD‐MCI). In the third trial, it is currently uncertain whether the intervention is best classified as cognitive training or cognitive rehabilitation (or possibly a combination of both of these approaches) due to limited information provided. These three trials are described in the Characteristics of studies awaiting classification table.

Excluded studies

We excluded a total of 47 studies (see Characteristics of excluded studies). Further information regarding these studies appears below.

RCTs of cognitive training or other interventions in PD‐MCI with no control group

Two studies evaluated cognitive training in people with PD‐MCI in which there was no control comparison group (Reuter 2012; Biundo 2015). Reuter 2012 tested the effects of cognitive training (targeting attention, executive function, and memory training) versus cognitive and transfer training versus cognitive, transfer, and psychomotor training. The study by Biundo 2015 evaluated cognitive training (RehaCom; attention, concentration, planning, and memory exercises) as a stand‐alone intervention versus cognitive training with non‐invasive brain stimulation. Mahmoud 2018 evaluated cognitive remediation therapy versus motor imagery training for people with PD and cognitive dysfunction diagnosed by a cognitive assessment on RehaCom; in this RCT there was no control group. Vlagsma 2020 examined the effects of computerised cognitive training for attention (CogniPlus) with a cognitive rehabilitation intervention as the comparison group in people with PD and executive dysfunction. Maggio 2018 tested the effects of virtual reality cognitive rehabilitation versus standard cognitive training in people with PD and mild to moderate cognitive impairment (MMSE from 11 to 26). An onoing RCT NCT03836963 is testing effects of cognitive and memory strategy training in veterans with PD‐MCI. Active comparators in this three‐arm RCT are cognitive and memory training as stand‐alone interventions.

Other cognition‐based interventions in people with PDD and PD‐MCI

Quayhagen 2000 evaluated a carer‐led cognitive stimulation intervention (incorporating memory/problem solving and fluency activities) in people with AD, cardiovascular dementia, or PDD. In this study no separate data for PDD were provided, and diagnosis of PDD was not made using clinical criteria. Hindle 2016 evaluated cognitive rehabilitation in people with PDD, and Farzana 2015 the effects of a home‐based cognitive stimulation intervention in people with PD and mild to moderate cognitive impairment, in which diagnosis of PDD and PD‐MCI was not reported. This study employed a pre‐post design. McCormick 2018 tested the feasibility and acceptability of individual cognitive stimulation therapy for people with PD‐related dementias. We found one ongoing RCT NCT03335150 of cognitive rehabilitation (CogSMART‐PD) for people with PD‐MCI versus a support group control intervention.

Non‐RCTs of cognitive training or other cognition‐based interventions in people with PD‐MCI

Naismith 2013 evaluated the effects of computerised cognitive training (based on individually tailored cognitive exercises) alongside education versus a waiting‐list control condition in a controlled trial design (non randomised) in a mixed sample of people with PD, of which some had no cognitive impairment and some had PD‐MCI. Stiver 2015 tested the efficacy of a mobile gaming engine aimed at reducing cognitive interference via a pre‐post design. Kim 2016 used a pre‐post design to test the effects of cognitive training (PD‐CoRE) in people with PD and executive dysfunction.

Cognitive training in people with PD without dementia or MCI

We identified nine RCTs of cognitive training in people with PD where people with dementia or MCI, or both, were excluded.

Zimmermann 2014 tested the effects of computer‐based cognitive training (versus a non‐cognition‐specific computer sports game; Nintendo Wii) in PD without cognitive impairment; Sammer 2006 examined the effects of working memory training targeting executive function (versus a control intervention) in people with PD and no cognitive impairment. Peña 2014 tested the effects of integrative structured cognitive training (REHACOP) (targeting attention, memory, processing speed, language, and executive function) versus occupational group activities in PD excluding people with dementia. One ongoing RCT NCT01469741 investigates effects of prospective memory training in PD but no dementia.

Edwards 2013 tested the effects of a self‐administered speed of information processing training excluding people with dementia, and Strouwen 2017 examined the effects of integrated versus consecutive dual task practice cognitive training aimed at increasing gait performance (in this study people with MMSE scores ≤ 24 were excluded). Pompeu 2012 evaluated the effects of a Wii‐based motor and cognitive training intervention (training attention and working memory) versus balance exercise therapy, in which people with dementia were excluded (MMSE ≥ 23). Piemonte 2016 tested a declarative memory training intervention aimed at improving gait and activities of daily living in people with PD with no cognitive impairment. We found one ongoing study NCT02922530 testing effectiveness of a mobile cognitive training intervention for depression and quality of life in people with PD without cognitive impairment. Valdés 2017 examined the effects of information processing training versus a delayed control group in people with PD without dementia or MCI (MMSE ≥ 24). Fernandez‐del‐Olmo 2018 examined the effects of cognitive training versus cognitive training with concurrent physical exercise in people with PD without cognitive impairment. We further excluded two ongoing studies; NCT03680170 and NCT04048122; which evaluate effects of web‐based working memory training versus low‐dose short‐term memory training in PD without cognitive impairment and the feasibility of cognitive rehabilitation (MC4PD strategy training) versus standard care for people with PD and subjective cognitive decline (MoCA < 21; primary outcome goal attainment).

Multicomponent interventions or cognitive interventions targeting other domains in people with PD without dementia or MCI

Peters 2012 is an ongoing study evaluating a multidisciplinary intervention incorporating exercise rehabilitation and cognitive and speech activities versus standard exercise in people with PD (excluding people with cognitive impairment). Monticone 2015 tested the effects of a multidisciplinary intervention incorporating task‐oriented activities, cognitive‐behavioural therapy, and occupational therapy (people with a MMSE < 24 were excluded). The study by Pompeu 2016 tests the effects of a physiotherapy guideline in people with PD and no dementia or MCI versus Microsoft Kinect games training (Kinect‐Adventures‐based training; KABT) on postural control, cognition, and quality of life. Walton 2016 evaluates the effects of a cognitive training intervention targeting executive function, processing speed, and attention, where freezing of gait is the primary outcome; cognition is not measured in this study. In an ongoing RCT ACTRN12617000634370, a cognitive‐plus‐exercise‐enrichment intervention is compared to standard care in people with PD and no cognitive impairment. Fellman 2018 examined the effects of cognitive training on working memory versus quiz training in people with PD and no cognitive impairment; in this RCT people with dementia were excluded. Goedeken 2018 is testing the effects of encoding strategy training versus verbal rehearsal (control group) for people with PD; in this RCT people with PDD are excluded. An ongoing trial NCT01156714 investigates the effects of exercise training versus computerised memory training versus combined exercise and motor training versus a comparison control group in people with PD and no cognitive impairment. The study by Motlagh 2017 compares cognitive training for freezing of gait to a control comparison group in people with PD and no cognitive impairment.

Non‐RCTs of cognitive training or other interventions targeting cognition in people with PD without dementia or MCI

Mirelman 2011 examined the feasibility of virtual reality training incorporating cognitive components in people with PD and no dementia; Disbrow 2012 tested the effects of computerised cognitive and motor training aimed at improving motor‐related executive function in PD without dementia, whereas Canning 2008 examined multiple task walking training incorporating cognitive activities. Both Atias 2015 and Adamski 2016 used pre‐post methodology to assess the feasibility of cognitive training in people with PD without dementia or MCI. Nombela 2011 tested the feasibility of Sudoku‐based cognitive training; this study neither commented on the cognitive level of the participants nor reported whether people with cognitive impairment were excluded.

We found three studies using pre‐post test methodology to evaluate the feasibility of cognitive training in people with PD and subjective cognitive complaints. Mohlman 2011 tested the feasibility of attention process training in PD and subjective cognitive complaints of working memory; Sinforiani 2004 tested the effects of cognitive training targeting multiple domains (attention, abstract reasoning, visuo‐spatial abilities, and motor training) in people with PD and mild cognitive deficits (defined by neuropsychological evaluation) using pre‐post methodology; people with severe cognitive impairment and dementia were excluded. An ongoing study NCT02826785 used pre‐post methodology to examine the effects of cognitive training in PD; participants reported at least one problem with their daily cognitive performance (but had no dementia). Both Díez‐Cirarda 2017 and Fearon 2017 employed a non‐RCT design: the feasibility of cognitive training was tested in Díez‐Cirarda 2017, and of a virtual reality‐based intervention combining motor and cognitive training in Fearon 2017. Both studies included people with PD and no cognitive impairment.

Risk of bias in included studies

Two review authors independently assessed risk of bias for all seven included studies (see Characteristics of included studies). None of the included studies met criteria for low risk of bias in all domains. See Figure 2 and Figure 3.

Figure 2

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

Figure 3

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

Allocation

For the domain of random sequence generation, we considered two studies to be at low risk (Lawrence 2018; Petrelli 2014), one study at high risk (Folkerts 2018), and four studies at unclear risk of bias due to insufficient information on how the random assignment was performed (Alloni 2018; Cerasa 2014; Costa 2014; París 2011). For the allocation concealment domain, we judged three studies to be at low risk, Folkerts 2018; Lawrence 2018; París 2011, and four studies to be at unclear risk of bias, again due to insufficient detail in the published reports (Alloni 2018; Cerasa 2014; Costa 2014; Petrelli 2014).

Blinding

We judged two studies to be at low risk of performance bias; in these studies both the intervention (cognitive training) and control (sham cognitive intervention) groups were not informed about their group assignment or the rationale behind their training and could therefore be considered blind to the intervention (Cerasa 2014; Costa 2014). We considered the risk of bias for the other five studies to be unclear for this domain. We judged four studies to be at low risk, Cerasa 2014; Costa 2014; París 2011; Petrelli 2014, one study to be at unclear risk, Alloni 2018, and the remaining two studies to be at high risk of detection bias, Folkerts 2018; Lawrence 2018, because they reported that not all assessors were blind to treatment allocation or that it was not possible to blind assessors.

Incomplete outcome data

We judged three of the seven included studies to be at unclear risk of attrition bias. One of these studies reported the number of participants who did not complete the study (Alloni 2018), but not the reasons for withdrawal. Two studies reported excluding participants who completed fewer than 75% of sessions from the analyses (París 2011 Petrelli 2014). We assessed the remaining four studies as at low risk of attrition bias.

Selective reporting

We found no evidence of selective reporting in any study, therefore we classified all studies as being at low risk of bias for this domain.

Other potential sources of bias

We did not identify any other sources of bias in the included studies.

Effects of interventions

See: Summary of findings for the main comparison Cognitive training compared to control intervention for cognition in PDD and PD‐MCI

Positive results favour cognitive training for all outcomes. A summary of findings and assessment of the certainty of the evidence is presented in the summary of findings Table for the main comparison.

Primary outcomes

Global cognition

The meta‐analysis on effects of cognitive training on global cognition at the end of the intervention period included six studies (Alloni 2018; Cerasa 2014; Folkerts 2018; Lawrence 2018; París 2011; Petrelli 2014) with 178 participants (Analysis 1.1). We found no clear evidence of a difference between cognitive training and control interventions. The result favoured cognitive training, but did not reach statistical significance (standardised mean difference (SMD) 0.28, 95% confidence interval (CI) −0.03 to 0.59; I² = 0%; low‐certainty evidence; Figure 4).

Figure 4

Forest plot of comparison: 1 Cognitive training versus control group, outcome: 1.1 Global cognition post‐treatment.

Executive function

We pooled five studies (Alloni 2018; Cerasa 2014; Costa 2014; Lawrence 2018; París 2011) to examine the effects of cognitive training on executive function measured immediately after the end of the intervention (Analysis 1.2). We found no evidence of a difference between cognitive training and control interventions (SMD 0.10, 95% CI −0.28 to 0.48; I² = 1%; 112 participants; low‐certainty evidence; Figure 5).

Figure 5

Forest plot of comparison: 1 Cognitive training versus control group, outcome: 1.2 Executive function post‐treatment.

Attention

Pooling data from five studies (Alloni 2018; Cerasa 2014; Lawrence 2018; París 2011; Petrelli 2014) showed that cognitive training was superior to control in effects on attention at the end of treatment, although the result was imprecise and compatible with a large or with very little effect (SMD 0.36, 95% CI 0.03 to 0.68; I² = 0%; 160 participants; low‐certainty evidence; Analysis 1.3; Figure 6). Imprecision was increased in a sensitivity analysis excluding Petrelli 2014, where fewer than 50% of participants had PD‐MCI, and the result was no longer statistically significant (SMD 0.41, 95% CI −0.01 to 0.83; I² = 0%; 4 studies; 95 participants; low‐certainty evidence; Analysis 2.1).

Figure 6

Forest plot of comparison: 1 Cognitive training versus control group, outcome: 1.3 Attention post‐treatment.

Verbal memory

Our analysis of effects on verbal memory showed that cognitive training was superior to control at the end of treatment, although this result was also imprecise and compatible with large or very small effects (SMD 0.37, 95% CI 0.04 to 0.69; I² = 0%; five studies (Alloni 2018; Cerasa 2014; Lawrence 2018; París 2011; Petrelli 2014); 160 participants; low‐certainty evidence; Analysis 1.4; Figure 7). In the sensitivity analysis excluding Petrelli 2014, there was no clear evidence of any effect (estimate smaller and more imprecise) (SMD 0.25, 95% CI −0.16 to 0.66; I² = 0%; 4 studies; 95 participants; low‐certainty evidence; Analysis 2.2).

Figure 7

Forest plot of comparison: 1 Cognitive training versus control group, outcome: 1.4 Verbal memory post‐treatment.

Visual processing

We found no evidence of an effect of cognitive training on visual processing ability at end of treatment (SMD 0.30, 95% CI −0.21 to 0.81; I² = 0%; three studies (Cerasa 2014; Lawrence 2018; París 2011); 64 participants; low‐certainty evidence; Analysis 1.5; Figure 8).

Figure 8

Forest plot of comparison: 1 Cognitive training versus control group, outcome: 1.5 Visual processing post‐treatment.

Secondary outcomes

Activities of daily living

We found no evidence of an effect of cognitive training on ADLs at the end of the intervention period (SMD 0.03, 95% CI −0.47 to 0.53; I² = 0%; three studies (Folkerts 2018; Lawrence 2018; París 2011); 67 participants; low‐certainty evidence; Analysis 1.6; Figure 9).

Figure 9

Forest plot of comparison: 1 Cognitive training versus control group, outcome: 1.6 Activities of daily living post‐treatment.

Quality of life

We found no evidence of an effect of cognitive training on quality of life immediately after the end of sessions (SMD −0.01, 95% CI −0.35 to 0.33; I² = 0%; five studies (Cerasa 2014; Folkerts 2018; Lawrence 2018; París 2011; Petrelli 2014) 147 participants; low‐certainty evidence; Analysis 1.7; Figure 10).

Figure 10

Forest plot of comparison: 1 Cognitive training versus control group, outcome: 1.7 Quality of life post‐treatment.

Adverse events

Folkerts 2018 reported no adverse events. The remaining studies did not provide any information about adverse events (Alloni 2018; Cerasa 2014; Costa 2014; Lawrence 2018; París 2011; Petrelli 2014).

Analyses on effects of cognitive training at longer‐term follow‐up (12 to 24 weeks)

Global cognition

The meta‐analysis on effects of cognitive training on global cognition at longer‐term follow‐up (12 to 24 weeks) included two studies (Alloni 2018; 24 weeks; Lawrence 2018; 12 weeks) with 41 participants. We found no evidence of a difference between groups (mean difference 0.28, 95% CI −1.73 to 2.28; I² = 0%; low‐certainty evidence; Analysis 1.8).

Executive function

We found no evidence of an effect of cognitive training on executive function at longer‐term follow‐up (SMD −0.22, 95% CI −0.85 to 0.41; I² = 0%; two studies (Alloni 2018; 24 weeks; Lawrence 2018; 12 weeks); 41 participants; low‐certainty evidence; Analysis 1.9).

Attention

We found no evidence of an effect of cognitive training on attention at longer‐term follow‐up (SMD 0.21, 95% CI −0.59 to 1.01; I² = 36%; two studies (Alloni 2018; 24 weeks; Lawrence 2018; 12 weeks) 41 participants; low‐certainty evidence; Analysis 1.10).

Verbal memory

We found no evidence of an effect of cognitive training on verbal memory at longer‐term follow‐up (SMD 0.15, 95% CI −0.47 to 0.78; I² = 0%; two studies (Alloni 2018; 24 weeks; Lawrence 2018; 12 weeks); 41 participants; low‐certainty evidence; Analysis 1.11).

Discussion

Summary of main results

The aim of this review was to evaluate the effectiveness of cognitive training interventions for people with PDD and PD‐MCI and to report on the quality of the evidence. Our results do not support a beneficial effect of cognitive training on general cognition shortly after the intervention has ended. Although cognitive training was associated with better scores on attention and verbal memory, the differences between the intervention and control groups were not robust to the exclusion of a study where only 20% of the sample had PD‐MCI. We found no evidence of benefit in other specific cognitive domains (executive function, visual processing ability) and no differences between cognitive training and control interventions in relation to ADLs and quality of life. Because most studies were in people with PD‐MCI who, by definition, have no significant functional impairment at baseline, effects on ADLs were unlikely to be detected. We considered the evidence to be of low certainty overall due to imprecision and study limitations (see Quality of the evidence).

Overall completeness and applicability of evidence

Only seven studies with a total of 225 participants met our inclusion criteria, therefore the evidence base remains very small. Only one study (and no identified ongoing studies) included people with PDD, reflecting an important gap in the evidence. Most of the included studies involved people with PD and mild cognitive impairment living in community settings. Uncertainty about the degree and nature of cognitive impairment in the study populations is an important limitation of our review. In one included study (Cerasa 2014), participants had significant cognitive impairment on a standard neuropsychological test, but their status against MCI criteria was unclear. We also included two studies in which the sample was a mixed population of people with and without MCI (París 2011; Petrelli 2014), as we were unable to access separate data. Given that diagnostic criteria for PD‐MCI have only recently been defined, and the variability in diagnosis (Marras 2013), we considered it important to be inclusive in our approach and to consider studies with mixed samples of people with and without clinically significant cognitive decline (as verified by performance on a standardised test) as eligible. As our sensitivity analyses showed that results were influenced by the inclusion of a study in which only a small proportion of the sample had cognitive impairment, it will be important that data on participants with and without cognitive impairment can be separated in future versions of this review or in other evidence syntheses.

Cognitive training across studies targeted either one or many cognitive domains. Single‐domain cognitive training focused on executive function in Alloni 2018, memory in Costa 2014 (prospective memory via training of shifting ability), and attention in Cerasa 2014. The remaining four studies evaluated cognitive training that targeted multiple domains of cognitive function (Folkerts 2018, executive function and visuo‐spatial ability; Lawrence 2018 and París 2011, attention, working memory, psychomotor speed, executive function, and visuo‐spatial ability; Petrelli 2014, attention, memory, and executive function, with additional psychoeducational elements). In all included studies cognitive training was computerised; in some studies this was augmented with paper and pencil homework assignments. To date we are unable to comment on whether multidomain or single‐domain cognitive training may differentially affect cognition in people with PD‐MCI and PDD.

Most of the included studies measured both general cognition and specific areas of cognitive function. Although outcomes were similar, studies used different measures, and in most studies primary outcome measures were not specified. It will be important that future research incorporates key core outcomes so that results across studies can be compared. We found only two studies reporting on long‐term effectiveness of cognitive training, measuring outcomes at 12 and 24 weeks. We found no evidence that cognitive training benefits global or specific areas of cognitive function in the longer term. However, studies were small, and risk of bias identified in several areas limits any conclusions about whether these interventions could benefit people with PDD and PD‐MCI at long‐term follow‐up. Overall, the current evidence therefore provides limited information in terms of potential long‐term benefit. Given the small number of studies and small sample sizes, our analyses may have had limited power to detect differences in cognition and in the secondary outcomes of ADLs and quality of life.

There was limited information on adverse events, with only one study reporting that the intervention was not associated with any adverse outcomes. This is in line with systematic reviews of cognitive training in older people and people with mild to moderate Alzheimer's disease and vascular dementia (Bahar‐Fuchs 2019), which found few studies mentioning adverse effects.

Although retrospectively it appears that we excluded many studies, we considered it important to include only studies where participants with PD also had clinical cognitive impairment to ensure that our review did not replicate previous reviews of effectiveness of cognitive training in PD without cognitive impairment (Leung 2015).

We were not able to conduct any subgroup analyses due to the small number of studies identified. An important aim for future versions of this review will be to assess whether the nature of the comparator intervention influences efficacy, as was found in a review of cognitive training in mild to moderate dementia (Bahar‐Fuchs 2019).

Quality of the evidence

All of the included studies were at unclear or high risk of bias in two or more domains. Many of the studies did not report details of random sequence generation, allocation concealment, or blinding. In some studies participants were excluded from analyses, indicating that intention‐to‐treat principles were not applied. For example, in two studies participants who did not receive more than 75% of the treatment were excluded from analyses.

According to the GRADE criteria, we considered the overall certainty of the evidence to be low for all outcomes due to risk of bias and imprecision. The numbers of studies and participants were small, and the confidence intervals around all effect estimates were wide. This GRADE rating means that new evidence may substantially change the effect estimates. Despite clinical heterogeneity across studies, such as inclusion of people with varying degrees of cognitive impairment, there was no or limited statistical heterogeneity across all of our analyses.

Potential biases in the review process

Although we searched systematically several sources, including the Cochrane Central Register of Controlled Trials (CENTRAL), conference proceedings, and review articles, the possibility remains that we may have missed some studies. We found several RCTs that reported excluding people with moderate to severe dementia; although these studies did not meet our inclusion criteria, it is possible that they included people with PD‐MCI and mild PDD.

Agreements and disagreements with other studies or reviews

Leung 2015 conducted a systematic review on the effectiveness of cognitive training for people with PD but no cognitive impairment. Although in some of our analyses cognitive training was favoured, which was in line with this review, our results did not reach statistical significance after sensitivity analyses were conducted. Given the small number of studies to date and risk of bias identified in several areas, our confidence in the conclusions of our review is limited.

Figure 1

Study flow diagram.

Navigate to figure in ReviewOpen in new tab

Figure 2

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

Navigate to figure in ReviewOpen in new tab

Figure 3

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

Navigate to figure in ReviewOpen in new tab

Figure 4

Forest plot of comparison: 1 Cognitive training versus control group, outcome: 1.1 Global cognition post‐treatment.

Navigate to figure in ReviewOpen in new tab

Figure 5

Forest plot of comparison: 1 Cognitive training versus control group, outcome: 1.2 Executive function post‐treatment.

Navigate to figure in ReviewOpen in new tab

Figure 6

Forest plot of comparison: 1 Cognitive training versus control group, outcome: 1.3 Attention post‐treatment.

Navigate to figure in ReviewOpen in new tab

Figure 7

Forest plot of comparison: 1 Cognitive training versus control group, outcome: 1.4 Verbal memory post‐treatment.

Navigate to figure in ReviewOpen in new tab

Figure 8

Forest plot of comparison: 1 Cognitive training versus control group, outcome: 1.5 Visual processing post‐treatment.

Navigate to figure in ReviewOpen in new tab

Figure 9

Forest plot of comparison: 1 Cognitive training versus control group, outcome: 1.6 Activities of daily living post‐treatment.

Navigate to figure in ReviewOpen in new tab

Figure 10

Forest plot of comparison: 1 Cognitive training versus control group, outcome: 1.7 Quality of life post‐treatment.

Navigate to figure in ReviewOpen in new tab

Analysis 1.1

Comparison 1 Cognitive training versus control group, Outcome 1 Global cognition post‐treatment.

Navigate to figure in ReviewOpen in new tab

Analysis 1.2

Comparison 1 Cognitive training versus control group, Outcome 2 Executive function post‐treatment.

Navigate to figure in ReviewOpen in new tab

Analysis 1.3

Comparison 1 Cognitive training versus control group, Outcome 3 Attention post‐treatment.

Navigate to figure in ReviewOpen in new tab

Analysis 1.4

Comparison 1 Cognitive training versus control group, Outcome 4 Verbal memory post‐treatment.

Navigate to figure in ReviewOpen in new tab

Analysis 1.5

Comparison 1 Cognitive training versus control group, Outcome 5 Visual processing post‐treatment.

Navigate to figure in ReviewOpen in new tab

Analysis 1.6

Comparison 1 Cognitive training versus control group, Outcome 6 Activities of daily living post‐treatment.

Navigate to figure in ReviewOpen in new tab

Analysis 1.7

Comparison 1 Cognitive training versus control group, Outcome 7 Quality of life post‐treatment.

Navigate to figure in ReviewOpen in new tab

Analysis 1.8

Comparison 1 Cognitive training versus control group, Outcome 8 Global cognition long term.

Navigate to figure in ReviewOpen in new tab

Analysis 1.9

Comparison 1 Cognitive training versus control group, Outcome 9 Executive function long term.

Navigate to figure in ReviewOpen in new tab

Analysis 1.10

Comparison 1 Cognitive training versus control group, Outcome 10 Attention long term.

Navigate to figure in ReviewOpen in new tab

Analysis 1.11

Comparison 1 Cognitive training versus control group, Outcome 11 Verbal memory long term.

Navigate to figure in ReviewOpen in new tab

Analysis 2.1

Comparison 2 Sensitivity analyses: cognitive training versus control group, Outcome 1 Attention post‐treatment.

Navigate to figure in ReviewOpen in new tab

Analysis 2.2

Comparison 2 Sensitivity analyses: cognitive training versus control group, Outcome 2 Verbal memory post‐treatment.

Navigate to figure in ReviewOpen in new tab

Summary of findings for the main comparison. Cognitive training compared to control intervention for cognition in PDD and PD‐MCI

Cognitive training compared to control intervention for cognition in PDD and PD‐MCI
Patient or population: cognition in PDD and PD‐MCI Setting: community and long‐term care Intervention: cognitive training Comparison: control intervention (no intervention, participating in recreational activities, receiving speech or language exercises, computerised motor therapy, or motor rehabilitation combined with recreational activity)
Outcomes	SMD (95% CI) meta‐analysis	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
Global cognition post‐treatment Assessed with: MMSE, CERAD Follow‐up: range 4 weeks to 8 weeks	SMD 0.28 higher (0.03 lower to 0.59 higher)	178 (6 RCTs)	⊕⊕⊝⊝ LOW ^{1 2}	A higher score is indicative of improved cognition.
Executive function post‐treatment Assessed with: Trail Making Test B, Stockings of Cambridge Follow‐up: range 4 weeks to 6 weeks	SMD 0.1 higher (0.28 lower to 0.48 higher)	112 (5 RCTs)	⊕⊕⊝⊝ LOW ^{1 2}	A higher score is indicative of improved executive function.
Attention post‐treatment Assessed with: Stroop Task and Brief Test of Attention Follow‐up: range 4 weeks to 6 weeks	SMD 0.36 higher (0.03 higher to 0.68 higher)	160 (5 RCTs)	⊕⊕⊝⊝ LOW ^{1 2}	A higher score is indicative of improved attention.
Verbal memory post‐treatment Assessed with: WMS Logical Memory test, Selective Reminding Test, Hopkins Verbal Learning Test‐Revised, Verbal short‐term memory DemTect Follow‐up: range 4 weeks to 6 weeks	SMD 0.37 higher (0.04 higher to 0.69 higher)	160 (5 RCTs)	⊕⊕⊝⊝ LOW ^{1 2}	A higher score is indicative of improved memory.
Visual processing post‐treatment Assessed with: Judgement Line Orientation Test Follow‐up: range 4 weeks to 6 weeks	SMD 0.3 higher (0.21 lower to 0.81 higher)	64 (3 RCTs)	⊕⊕⊝⊝ LOW ^{1 2}	A higher score is indicative of improved visual processing.
Activities of daily living post‐treatment Assessed with: Barthel Index, Unified Parkinson's Disease Rating Scale, Cognitive Difficulties Scale Follow‐up: range 4 weeks to 8 weeks	SMD 0.03 higher (0.47 lower to 0.53 higher)	67 (3 RCTs)	⊕⊕⊝⊝ LOW ^{1 2}	A higher score is indicative of improved activities of daily living.
Quality of life post‐treatment Assessed with: Parkinson's Disease Questionnaire, QUALIDEM Follow‐up: range 4 weeks to 8 weeks	SMD 0.01 lower (0.35 lower to 0.33 higher)	147 (5 RCTs)	⊕⊕⊝⊝ LOW ^{1 2}	A higher score is indicative of improved quality of life.
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RCT: randomised controlled trial; SMD: standardised mean difference
GRADE Working Group grades of evidence High certainty: We are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: We are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: Our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: We have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.
¹Downgraded one point due to risk of bias (all studies have at least two domains at unclear risk of bias with none of the studies at low risk of bias in all domains). ²Downgraded one point for imprecision due to small sample size (< 400 participants) and wide confidence intervals. PDD: Parkinson's disease dementia PD‐MCI: Parkinson's Disease–Mild Cognitive Impairment MMSE: Mini‐Mental State Examination CERAD: Consortium to Establish a Registry for Alzheimer's Disease test battery WMS: Wechsler Memory Scale

Summary of findings for the main comparison. Cognitive training compared to control intervention for cognition in PDD and PD‐MCI

Navigate to table in Review

Comparison 1. Cognitive training versus control group

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Global cognition post‐treatment Show forest plot	6	178	Std. Mean Difference (IV, Random, 95% CI)	0.28 [‐0.03, 0.59]

2 Executive function post‐treatment Show forest plot	5	112	Std. Mean Difference (IV, Random, 95% CI)	0.10 [‐0.28, 0.48]

3 Attention post‐treatment Show forest plot	5	160	Std. Mean Difference (IV, Random, 95% CI)	0.36 [0.03, 0.68]

4 Verbal memory post‐treatment Show forest plot	5	160	Std. Mean Difference (IV, Random, 95% CI)	0.37 [0.04, 0.69]

5 Visual processing post‐treatment Show forest plot	3	64	Std. Mean Difference (IV, Random, 95% CI)	0.30 [‐0.21, 0.81]

6 Activities of daily living post‐treatment Show forest plot	3	67	Std. Mean Difference (IV, Random, 95% CI)	0.03 [‐0.47, 0.53]

7 Quality of life post‐treatment Show forest plot	5	147	Std. Mean Difference (IV, Random, 95% CI)	‐0.01 [‐0.35, 0.33]

8 Global cognition long term Show forest plot	2	41	Mean Difference (IV, Random, 95% CI)	0.28 [‐1.73, 2.28]

9 Executive function long term Show forest plot	2	41	Std. Mean Difference (IV, Random, 95% CI)	‐0.22 [‐0.85, 0.41]

10 Attention long term Show forest plot	2	41	Std. Mean Difference (IV, Random, 95% CI)	0.21 [‐0.59, 1.01]

11 Verbal memory long term Show forest plot	2	41	Std. Mean Difference (IV, Random, 95% CI)	0.15 [‐0.47, 0.78]

Comparison 1. Cognitive training versus control group

Navigate to table in Review

Comparison 2. Sensitivity analyses: cognitive training versus control group

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Attention post‐treatment Show forest plot	4	95	Std. Mean Difference (IV, Random, 95% CI)	0.41 [‐0.01, 0.83]

2 Verbal memory post‐treatment Show forest plot	4	95	Std. Mean Difference (IV, Random, 95% CI)	0.25 [‐0.16, 0.66]

Comparison 2. Sensitivity analyses: cognitive training versus control group

Navigate to table in Review

Cochrane Review language

Website language

Resumen

Antecedentes

Objetivos

Métodos de búsqueda

Criterios de selección

Obtención y análisis de los datos

Resultados principales

Conclusiones de los autores

PICOs

PICOs

Population

Intervention

Comparison

Outcome

Resumen en términos sencillos

Intervenciones de entrenamiento cognitivo para la demencia y el deterioro cognitivo leve en la enfermedad de Parkinson

Visual summary

Authors' conclusions

Implications for practice

Implications for research

Summary of findings

Background

Description of the condition

Description of the intervention

How the intervention might work

Why it is important to do this review

Objectives

Primary objective

Secondary objectives

Methods

Criteria for considering studies for this review

Types of studies

Types of participants

Types of interventions

Types of outcome measures

Primary outcomes

Secondary outcomes

Search methods for identification of studies

Electronic searches

Searching other resources

Data collection and analysis

Selection of studies

Data extraction and management

Assessment of risk of bias in included studies

Measures of treatment effect

Unit of analysis issues

Dealing with missing data

Assessment of heterogeneity

Assessment of reporting biases

Data synthesis

Subgroup analysis and investigation of heterogeneity

Sensitivity analysis

Summary of findings and assessment of the certainty of the evidence

Results

Description of studies

Results of the search

Included studies

Design

Setting

Participants

Cognitive training interventions and control comparisons

Adherence

Outcomes

Primary outcome ‐ Cognition

Secondary outcomes

Ongoing trials

Studies awaiting classification

Excluded studies

RCTs of cognitive training or other interventions in PD‐MCI with no control group

Other cognition‐based interventions in people with PDD and PD‐MCI

Non‐RCTs of cognitive training or other cognition‐based interventions in people with PD‐MCI

Cognitive training in people with PD without dementia or MCI

Multicomponent interventions or cognitive interventions targeting other domains in people with PD without dementia or MCI

Non‐RCTs of cognitive training or other interventions targeting cognition in people with PD without dementia or MCI

Risk of bias in included studies

Allocation

Blinding

Incomplete outcome data