Cognitive behavioural therapy (CBT) with and without exercise to reduce fear of falling in older people living in the community

Abstract

Background

Fear of falling (FoF) is a lasting concern about falling that leads to an individual avoiding activities that he/she remains capable of performing. It is a common condition amongst older adults and may occur independently of previous falls. Cognitive behavioural therapy (CBT), a talking therapy that helps change dysfunctional thoughts and behaviour, with and without exercise, may reduce FoF, for example, by reducing catastrophic thoughts related to falls, and modifying dysfunctional behaviour.

Objectives

To assess the benefits and harms of CBT for reducing FoF in older people living in the community, and to assess the effects of interventions where CBT is used in combination with exercise.

Search methods

We searched the Cochrane Central Register of Controlled Trials (CENTRAL, Issue 1, 2023), MEDLINE Ovid (from 1946 to 11 January 2023), Embase Ovid (from 1980 to 11 January 2023), CINAHL Plus (Cumulative Index to Nursing and Allied Health Literature) (from 1982 to 11 January 2023), PsycINFO (from 1967 to 11 January 2023), and AMED (Allied and Complementary Medicine from 1985 to 11 January 2023). We handsearched reference lists and consulted experts for identifying additional studies.

Selection criteria

This review included randomised controlled trials (RCTs), quasi‐RCTs, and cluster‐RCTs assessing CBT with and without exercise interventions compared to control groups with sham‐treatment, or treatment as usual. We defined CBT as a collaborative, time‐limited, goal‐oriented, and structured form of speaking therapy. Included studies recruited community‐dwelling older adults, with a mean population age of at least 60 years minus one standard deviation, and not defined by a specific medical condition.

Data collection and analysis

Two review authors used standard methodological procedures expected by Cochrane. For continuous data, as assessed by single‐ or multiple‐item questionnaires, we report the mean difference (MD) with 95% confidence interval (CI) when studies used the same outcome measures, and standardised mean difference (SMD) when studies used different measures for the same clinical outcome. For dichotomous outcomes, we reported the treatment effects as risk ratios (RR) with 95% CIs. We measured the primary outcome, FoF, immediately, up to, and more than six months after the intervention. We analysed secondary outcomes of activity avoidance, occurrence of falls, depression, and quality of life when measured immediately after the intervention. We assessed risk of bias for each included study, using the GRADE approach to assess the certainty of evidence.

Main results

We selected 12 studies for this review, with 11 studies included for quantitative synthesis. One study could not be included due to missing information. Of the 11 individual studies, two studies provided two comparisons, which resulted in 13 comparisons. Eight studies were RCTs, and four studies were cluster‐RCTs. Two studies had multiple arms (CBT only and CBT with exercise) that fulfilled the inclusion criteria. The primary aim of 10 studies was to reduce FoF. The 11 included studies for quantitative synthesis involved 2357 participants, with mean ages between 73 and 83 years. Study total sample sizes varied from 42 to 540 participants. Of the 13 comparisons, three investigated CBT‐only interventions while 10 investigated CBT with exercise. Intervention duration varied between six and 156 hours, at a frequency between three times a week and monthly over an eight‐ to 48‐week period. Most interventions were delivered in groups of between five and 10 participants, and, in one study, up to 25 participants. Included studies had considerable heterogeneity, used different questionnaires, and had high risks of bias.

CBT interventions with and without exercise probably improve FoF immediately after the intervention (SMD −0.23, 95% CI −0.36 to −0.11; 11 studies, 2357 participants; moderate‐certainty evidence). The sensitivity analyses did not change the intervention effect significantly. Effects of CBT with or without exercise on FoF may be sustained up to six months after the intervention (SMD −0.24, 95% CI −0.41 to −0.07; 8 studies, 1784 participants; very low‐certainty evidence). CBT with or without exercise interventions for FoF probably sustains improvements beyond six months (SMD −0.28, 95% CI −0.40 to −0.15; 5 studies, 1185 participants; moderate‐certainty of evidence).

CBT interventions for reducing FoF may reduce activity avoidance (MD −2.57, 95% CI −4.67 to −0.47; 1 study, 312 participants; low‐certainty evidence), and level of depression (SMD −0.41, 95% CI −0.60 to −0.21; 2 studies, 404 participants; low‐certainty evidence). We are uncertain whether CBT interventions reduce the occurrence of falls (RR 0.96, 95% CI 0.66 to 1.39; 5 studies, 1119 participants; very low‐certainty evidence).

All studies had a serious risk of bias, due to performance bias, and at least an unclear risk of detection bias, as participants and assessors could not be blinded due to the nature of the intervention. Downgrading of certainty of evidence also occurred due to heterogeneity between studies, and imprecision, owing to limited sample size of some studies. There was no reporting bias suspected for any article.

No studies reported adverse effects due to their interventions.

Authors' conclusions

CBT with and without exercise interventions probably reduces FoF in older people living in the community immediately after the intervention (moderate‐certainty evidence). The improvements may be sustained during the period up to six months after intervention (low‐certainty evidence), and probably are sustained beyond six months (moderate‐certainty evidence). Further studies are needed to improve the certainty of evidence for sustainability of FoF effects up to six months.

Of the secondary outcomes, we are uncertain whether CBT interventions for FoF reduce the occurrence of falls (very low‐certainty evidence). However, CBT interventions for reducing FoF may reduce the level of activity avoidance, and may reduce depression (low‐certainty evidence). No studies reported adverse effects.

Future studies could investigate different populations (e.g. nursing home residents or people with comorbidities), intervention characteristics (e.g. duration), or comparisons (e.g. CBT versus exercise), investigate adverse effects of the interventions, and add outcomes (e.g. gait analysis). Future systematic reviews could search specifically for secondary outcomes.

Plain language summary

Cognitive behavioural interventions for reducing fear of falling in older people living in the community

Key messages

– Cognitive behavioural therapy (CBT) with and without exercise probably reduces fear of falling in older people living in the community, when measured once treatment has ended. Improvements may be sustained during the first six months after treatment finished, and probably last beyond six months.

– As a consequence of these interventions, people may be less likely to avoid activities after treatment and their level of depression may be reduced.

– It is unclear if the frequency of falls is reduced following treatment.

– We do not know if there are any adverse effects (harms) caused by CBT with and without exercise for reducing fear of falling, as none of the studies measured this as one of their outcomes. We need more studies looking at adverse effects.

What is fear of falling?

Fear of falling is a lasting concern about falling that leads to a person avoiding activities that he/she remains capable of performing. Fear of falling is common among older adults. They may be warned by healthcare professionals, family, and friends of the dangers of falls, as well have witnessed directly or indirectly the consequences of a fall. This is significant as up to 34% of older adults fall each year, with 5% experiencing bone fractures. Furthermore, they may recognise that their body is not as strong as it was when they were younger, adding to concern that they may not be able to protect themselves from a fall, and must, therefore, take preventive measures to avoid falling. People with fear of falling can experience physical, psychological, and social consequences. Treating fear of falling is therefore important to reduce dysfunctional cognitions and behaviours leading to these consequences.

How is the condition treated?

There are several treatment approaches: cognitive behavioural therapy (CBT) (a talking therapy that helps change thoughts and behaviour), exercise (a planned, structured, and repetitive physical activity to help keep the body healthy), or a combination of both. These treatments are usually given in group settings by trained therapists.

What did we want to find out?

We wanted to find out if CBT with and without exercise in older adults living in the community (who live in places without additional support, such as assisted living centres) were better than usual care or dummy treatments in reducing fear of falling. We also wanted to see how CBT with and without exercise affected activity avoidance, falls, and depression, or if it caused any harm.

What did we do? 
We searched several electronic databases and consulted experts for studies that compared interventions to reduce fear of falling using CBT alone and CBT with exercise.

We combined and summarised the results across the studies. We rated our confidence in the evidence based on factors such as study design, methods, and numbers of participants.

What did we find?

We found 12 relevant studies, of which 11 studies were included for statistical analyses with a total of 2383 people, with a mean age varying from 73 to 83 years. The therapy (CBT or dummy treatment) was given at a frequency from three times per week to once per month, for eight to 48 weeks. Added up, the treatments lasted between six and 156 hours. Most interventions were given in groups of between five and 10 participants, and in one study up to 25. The primary aim of 10 studies was to reduce fear of falling.

Main results

We found that CBT with and without exercise interventions probably reduces fear of falling in older people living in the community once treatment has ended. Improvements may be sustained during the first six months after treatment finished, and probably last beyond six months. Additionally, we found that people may be less likely to avoid activities, and may reduce their level of depression. It remains unclear if the frequency of falls improves after treatment.

What are the limitations of the evidence?

Our confidence in the evidence was limited because the results may have been influenced by the participants in the studies knowing which treatment they received and the studies used different ways of delivering the interventions.

To improve our certainty of the evidence, we would need more studies, with more similarity in how they treated and measured fear of falling.

How up to date is this evidence?

This evidence is up to date to 11 January 2023.

Summary of findings

Summary of findings 1. CBT interventions (CBT with and without exercise) compared to control (standard care or sham treatments) for reducing fear of falling (FoF) in older people living in the community.

CBT interventions (with and without exercise) compared to control (standard care or sham treatments) for reducing fear of falling (FoF) in older people living in the community
Patient or population: older adults (mean age minus 1 standard deviation is more than 60 years) Setting: community living Intervention: CBT interventions (with and without exercise) Comparison: control (standard care or sham treatments) for FoF
Outcomes	Anticipated absolute effects* (95% CI)		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Risk with control (standard care or sham treatments) for FoF	Risk with CBT interventions (CBT with and without exercise)
FoF (immediate postintervention)a Assessed with: FES‐S, mFES, FES‐I, FES, ABC, and single‐item instruments Follow‐up: immediately after intervention	—	SMD 0.23 lower (0.36 lower to 0.11 lower)b	—	2357 (11 RCTs)c	⊕⊕⊕⊝ Moderated	This SMD indicated a small effect.b This corresponds to a reduction on the FES‐I scale of 2.2 (95% CI −3.4 to −1.0). In the control group, mean FES‐S score was 106.2, mFES ranged from 28.2 to 35.3, mean FES‐I from 27.15 to 39.53, mean FES was 88.69, ABC was 150.3, mean single‐item instruments from 1.7 to 1.8.
FoF (sustainability of effects, up to 6 months' postintervention) Assessed with: FES‐S, mFES, FES‐I, FES, and ABC Follow‐up: 1–6 months after the intervention	—	SMD 0.24 lower (0.41 lower to 0.07 lower)b	—	1784 (8 RCTs)e	⊕⊕⊝⊝ Lowd,f	This SMD indicated a small effect.b This corresponds to a reduction on the FES‐I scale of 2.3 (95% CI −4.3 to −0.19). In the control group, mean FES‐S score was 99.0, mFES ranged from 29.4 to 35.86, mean FES‐I from 34.1 to 38.07, mean FES was 88.4, ABC was 146.9.
FoF (sustainability of effects, > 6 months' postintervention) Assessed with: FES‐S, mFES, and FES‐I Follow‐up: 12–24 months' postintervention	—	SMD 0.28 lower (0.40 lower to 0.15 lower) b	—	1185 (5 RCTs)	⊕⊕⊕⊝ Moderated	This SMD indicated a small effect.b This corresponds to a reduction on the FES‐I scale of 2.6 (95% CI −3.8 to −1.8). In the control group, mean FES‐S score was 99.0, mFES ranged from 28.9 to 35.86, mean FES‐I mean was 38.68.
Activity avoidance Assessed with: FES‐IAB Follow‐up: immediately after intervention	In the control group, mean FES‐IAB score was 28.74	MD 2.57 lower (4.67 lower to 0.47 lower)	—	312 (1 RCT)	⊕⊕⊝⊝ Lowd,g	—
Occurrence of falls Follow‐up: immediately after intervention	482 per 1000h	462 per 1000 (318 to 669)	RR 0.96 (0.66 to 1.39)	1119 (5 RCTs)	⊕⊝⊝⊝ Very lowd,f,g	An RR < 1 favours intervention group.
Depression Assessed with: HADS‐D and GDS Follow‐up: immediately after intervention	For CBT only interventions control group, mean HADS‐D score was 0.1	For CBT only interventions, MD 1.26 lower (1.96 lower to 0.56 lower)	—	For CBT only interventions, 314 (1 RCT)	⊕⊕⊝⊝ Lowd,g	SMD −0.41, 95% CI −0.60 to −0.21b
	For CBT with exercise interventions control group, mean GDS score was 0.79	For CBT with exercise interventions, MD 0.38 lower (0.80 lower to 0.04 higher)		For CBT with exercise interventions, 90 (1 RCT)
Adverse effects	—	—	—	—	—	No studies reported this outcome.
*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). ABC: Activities‐Specific Balance Confidence Scale, range 0 to 100, a high score represents a high level of physical functioning; CBT: cognitive behavioural therapy; CI: confidence Interval; FES: Falls Efficacy Scale, range 10 to 100, lower scores represent lower concerns about falling; FES‐I: FES‐International, range 16 to 64, lower scores represent lower concerns about falling; FES‐IAB: FES International Avoidance Behaviour, range 16 to 64, a low score represents low activity avoidance; FES‐S: Falls Efficacy Scale Swedish variant, range 0 to 130, lower scores represent lower concerns about falling; FoF: fear of falling; GDS: Geriatric Depression Scale, range 0 to 15, a low score represents low level of depression; HADS‐D: Hospital Anxiety and Depression Scale ‐ Depression subscale, range 0 to 21, a low score represents low level of depression; MD: mean difference; mFES: Modified FES, range 16 to 64, lower scores represent lower concerns about falling; RCT: randomised controlled trial; RR: risk ratio; Single‐Item Instruments: two instruments, ranging 1 to 4 and 1 to 5, a high score represents high FoF; 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.

^a The interventions fell into two categories: cognitive behavioural therapy (CBT) only interventions and multicomponent interventions, combining CBT and exercise interventions. Two studies were multiple arm studies, having both a CBT only arm and a CBT with exercise arm.
^b The rating of effect was based on a rule of thumb interpretation in which an SMD of 0.2 to 0.5 indicates a small effect, 0.5 to 0.8 a moderate effect, and over 0.8 a large effect.
^c There were 11 studies included for quantitative synthesis, of which two had two arms (CBT only and CBT with exercise) (Huang 2011; Reinsch 1992).
^d Downgraded one level due to serious risk of bias; blinding of participants and assessors not possible due to nature of the intervention.
^e There were eight studies included for quantitative synthesis, of which one had two arms (CBT only and CBT with exercise) (Huang 2011).
^f Downgraded one level due to serious inconsistency; heterogeneity existed between studies in CBT with exercise group; or heterogeneity existed between pooled studies.
^g Downgraded one level due to serious imprecision; CIs indicated both benefit and harm; or insufficient number of participants (fewer than 200).
^h Derived from the pooled estimate of the control group.

Background

Description of the condition

First introduced as a concept in 1982 by Murphy and Isaac, fear of falling (FoF) was defined in 1993 by Tinetti and colleagues as "a lasting concern about falling that leads to an individual avoiding activities that he/she remains capable of performing" (Murphy 1982; Tinetti 1993). FoF is an umbrella term relating to different concepts, assessment, and emerging intervention strategies. FoF, concern about falling, falls‐efficacy, fall‐related psychological concerns (FRCPs), ptophobia, fall‐related anxiety, and fear‐related activity avoidance are often used interchangeably; however, they represent different psychological constructs (Ellmers 2023; Jørstad 2005). The fear in FoF describes a state of apprehension towards a danger, the fall; "falls‐efficacy" describes the level of confidence towards the ability to mitigate the threat, that of falls, and as such is considered a resiliency factor influencing the level of FoF (Payette 2016); "Concern about falling," is a term that is closely related to fear, but is less intense and emotional (and therefore may be more socially acceptable for older people to disclose) (Jørstad 2005; Yardley 2005). However, concern describes a solution‐orientated cognitive response whereas fear is an innate response (Levy 1985). Anxiety, often used to describe FoF, describes a state of apprehension towards a future, imagined danger.

The behavioural reactions seen in those with FoF may be maladaptive or adaptive (Ellmers 2022). Maladaptive behaviours seen in FoF reflects a mismatch between physiological fall risk and one's perceived risk of falling. Adamczewska 2018 proposes that FoF may have a protective element against falls, when there is a focus on the immediate threat, as felt by fear, rather than the future threat, as felt by anxiety, which would predispose to a vicious cycle of avoidance and a sedentary lifestyle. Therefore, FoF may be protective when there is low anxiety, and maladaptive when there is high anxiety. Delbaere 2010 suggested a classification based on this mismatch; low or high physiological fall risk, as measured by the physiological profile assessment (Lord 2003), and low FoF (Falls Efficacy Scale – International (FES‐I) score of 22 or less for low physiological fall risk or 19 or less for high physiological fall risk), or high FoF. The four possible states are classified as vigorous (low FoF and low fall risk), anxious (high FoF and low fall risk), stoic (low FoF and high fall risk), and aware (high FoF and high fall risk) (Delbaere 2010).

FoF is common amongst older people, usually defined as aged 65 years and above. Studies in community‐dwelling older adults show a FoF prevalence in both genders of 36%, 43% in women, and 26% in men (Boyd 2009; Tomita 2018), which increased with age (Lach 2005), and after falls (Lavedán 2018). One cross‐sectional study found a prevalence of 43.8% of FoF of people who had not fallen, 66% who had fallen once, and up to 86% in people who had fallen more than once (Zijlstra 2007). Considering that global yearly prevalence of falls, as shown in one recent meta‐analysis, in older adults ranged from 28% to 34%, of which 5% may end in fractures, there is likely a significant and growing group of people who experience falls (Salari 2022). Clemson 2015 found that having an injurious fall does not predict acquiring FoF and that having FoF did not predict having an injurious fall; however, FoF has been shown to predict falls when mobility was limited (Litwin 2018).

Risk factors associated with FoF are female gender, older age, lower level of education, chronic illness, poor subjective health status, functional impairments, the use of a walking aid, a history of falls, and depression (Asai 2022; Denkinger 2015; Park 2017). Anxiety was found to be an indirect predisposing factor for FoF, affecting self‐confidence at avoiding falls and restricting activity, which then predisposes to FoF (Denkinger 2015).

Protective factors against FoF have been found to be social support, participating in frequent social activities, increased opportunities for physical activities (Howland 1998; Lee 2018). Howland 1998 also reported that "those who could rely on others or talk with friends about falling were least likely to report activity curtailment".

FoF has serious adverse health outcomes. FoF is associated with avoidance of activity (Zijlstra 2007), decreased social participation (Pin 2016), lower self‐efficacy (Cumming 2000; Denkinger 2010), decreased ability to perform activities of daily living (ADLs) (Cumming 2000), a decreased quality of life (Schoene 2019), fear and anxiety (Painter 2012), and loss of balance confidence (Hadjistavropoulos 2012). Altered gait parameters have been associated with FoF, with a slowed gait, shorter stride length, slower step rate, longer double support time (time when both feet are on the ground), and greater variability in stride length (Makino 2017).

Secondary diagnoses, such as depression can also develop (Denkinger 2015). Consequently, the resultant health costs are significant (Heinrich 2010). Van Haastegt and colleagues found that FoF interventions could be more cost‐effective than treating the associated symptoms that may develop from FoF (van Haastregt 2013).

Measuring FoF has been achieved through different approaches (Greenberg 2012; McGarrigle 2023; Moore 2008; Soh 2022). Single‐item instruments involve direct questioning with yes/no questionnaires, for example, "are you afraid of falling?" (Tinetti 1990) or with Likert scales, for example, "How afraid are you that you will fall and hurt yourself in the next year?" (Lachman 1998). Alternatively, multi‐item indirect instruments to measure FoF do so through measuring confidence, in particular falls‐related self‐efficacy (Moore 2008). Falls‐related self‐efficacy in the context of FoF is the confidence to perform an activity without falling. Multi‐item instruments to assess falls‐related self‐efficacy have been found to be more reliable than single‐item instruments (Scheffer 2008). Multi‐item instruments are superior to dichotomous outcomes, allowing the assessment of FoF in both simple and complex activities in more detail and as a continuum, describing a level of FoF. In using single‐item questions or multi‐item questionnaires using Likert scaling, FoF can be followed over time to better identify changes. Questionnaires were developed based on different concepts of FoF, for example, targeting concerns of falling in the FES‐I (Yardley 2005), or balance confidence as in the Activities Specific Balance and Confidence Scale (ABC) (Powell 1995), representing the most established assessments strategies (Scheffer 2008).

Description of the intervention

The main therapeutic approaches to treat FoF are currently represented by exercise interventions and cognitive behavioural therapy (CBT). Exercise interventions are planned, structured, repetitive, and purposeful physical activities to improve or maintain one or more components of physical fitness (CDC 2020; Kendrick 2014), also targeting related health outcomes such as prevention of falls.

CBT interventions, specifically designed to treat FoF, represents a second approach to reduce FoF, and have been positively evaluated in one review (Liu 2019). CBT is a class of psychotherapeutic interventions based on the cognitive theory, where an individual's idiosyncratic and maladaptive ideation, often called dysfunctional cognitions, and consequently dysfunctional behaviour, are modified, in a process called cognitive restructuring (Beck 1970). CBT interventions for FoF are psychotherapeutic interventions aimed at modifying individuals' thoughts and behaviour. This therapeutic approach is to restructure maladaptive behaviour to a more adaptive one (Liu 2019; Tennstedt 1998). CBT is collaborative, time‐limited, goal‐oriented, and structured. Misconceptions, distortions, and maladaptive assumptions are identified, and the validity and reasonableness of the assumptions tested (Beck 1970). Interventions aim to lower the emotional and behavioural sequela resulting from these dysfunctional cognitions, by loosening the perseverative and distorted cognitions, and by introducing more realistic ones (Beck 1970). CBT is the most researched form of psychotherapy, with evidence of efficacy in multiple psychiatric pathologies (Hofmann 2012). CBT interventions specifically designed for treating FoF focus on cognitive restructuring, goal setting, and promoting physical activities in both one‐on‐one or group sessions and are given by specially trained individuals (Liu 2019).

Another approach to treating FoF is a multicomponent intervention, that is, CBT with exercise interventions, where the techniques of these individual therapeutic approaches are used together (Huang 2016; Wetherell 2018).

The first widely implemented CBT with exercise intervention developed for treating FoF was Tennstedt 1998's "A Matter of Balance" (AMB) and has been adapted to different languages and countries. This time‐limited intervention is given in twice‐weekly group sessions, starting with psychotherapeutic discussions, progressing onto physical exercise, and ending with another discussion. A cognitive restructuring approach to changing behaviour is the main therapeutic approach (Tennstedt 1998). The group discussions are focused and have a new theme for each session. The exercise consists of strength training exercises.

How the intervention might work

Several mechanisms may be involved in reducing FoF in both CBT, and CBT with exercise. CBT for FoF aims at modifying dysfunctional cognitions regarding the risk of falling, and the resulting behaviour. The specific cognitive mechanisms maintaining FoF have been linked to those of post‐traumatic stress disorder (PTSD), where there is a trauma response involving anxiety, loss of self‐confidence, and activity avoidance (Adamczewska 2018). A Multifactorial Causation Model of Falls and Fear has been proposed linking FoF with appraisals of an individual's ability to maintain balance in relation to other contributors such as falls history or beliefs that a person holds regarding falls (Hadjistavropoulos 2012). Such an extended model would also help explain the high frequency of FoF in those who have not had a severe fall or any fall in the recent past. The Multifactorial Causation Model of Falls and Fear suggests that falls efficacy, the belief that one can avoid falling, has a role in stressful situations following exposure to fall trauma (Adamczewska 2018; Tinetti 1990). The beliefs towards falling therefore influence FoF.

Kendrick 2014 suggests that exercise may reduce FoF by directly impacting FoF or indirectly impacting factors associated with FoF and the risk of falling. The exercise component in CBT with exercise interventions is expected to have the same benefits. Exercise is expected to reduce the rate of falls through strength, balance, endurance, and stability training, as well as improving mood (Jung 2008; Kendrick 2014). Exercise may also improve fall‐related self‐efficacy (Kendrick 2014).

Li 2005 hypothesised that the mediator for change in FoF is fall‐related self‐efficacy. In following Bandura 2004's social cognitive theory, falls efficacy would mediate a decrease in FoF through the potentially pleasurable experience of exercise, the social approval of pursuing exercise, and the positive self‐evaluations of having successfully completed the exercise. Those with high self‐efficacy would therefore likely have low FoF.

In combination with CBT approaches, the exercise component of CBT with exercise interventions may serve as a source of behavioural experiments, where, through exposure between sessions to situations that evoke FoF, the individual acquires experience refuting prior beliefs, or dysfunctional cognitions, relating to falling, through performing activities that do not result in falls as expected. In considering the PTSD cognitive model, behavioural experiments during the exercise component of CBT with exercise interventions would reduce FoF by modifying negative appraisals of the trauma and its sequelae; elaborate the fall‐related traumatic memories into context of time, place, and situation; and drop dysfunctional behaviours and cognitive strategies (Ehlers 2005).

Why it is important to do this review

The high prevalence and associated serious adverse health outcomes make FoF in older adults a major public health issue (Boyd 2009). FoF is an interdisciplinary challenge encompassing both physical and mental health: primary care providers, such as geriatricians or family physicians, would manage mainly the physical sequelae, whereas mental health providers, such as psychiatrists, would respond to the psychological responses to FoF and the resulting vicious cycles leading to psychiatric comorbidity (Lenouvel 2021). Best available evidence for the effectiveness of treatment addressing both the physical and mental health needs of individuals with FoF is required.

In addition to updating the literature search, our review investigated other outcomes, including activity avoidance, occurrence of falls, depression, anxiety, and adverse outcomes, aiming to provide a comprehensive overview of the effects of CBT or CBT and exercise together, as these outcomes have not been included in current reviews of CBT interventions for reducing FoF. These outcomes are selected as they are known to be related to FoF.

We will also consider the interventions in the context of the need for a standardised taxonomy to characterise the interventions and to address problems of intervention heterogeneity (variability) in both research and care for falling (Lamb 2011).

Our review will complement the Kendrick 2014 Cochrane Review on exercise interventions to reduce FoF, as well as establishing the evidence base of the emerging CBT and combined CBT with exercise approaches. The Cochrane format of ongoing reviewing will allow inclusion of emerging new studies to allow an update of evidence found (Moher 2006).

Objectives

To assess the benefits and harms of CBT for reducing FoF in older people living in the community, and to assess the effects of interventions where CBT is used in combination with exercise.

Methods

Criteria for considering studies for this review

Types of studies

We searched for randomised controlled trials (RCTs), quasi‐RCTs (i.e. group allocation based on certain criteria), and cluster‐RCTs (if more than one cluster per intervention) assessing CBT with and without exercise interventions for FoF versus control.

Types of participants

Trials were included when all the participants were:

• community‐dwelling older adults. Community living was considered when the places of residence did not provide significant health‐related care, such as living at home, retirement villages, or shelters. Trials with mixed populations were included. Mixed populations are when people living in the community are grouped with people living in places where increased care is provided, such as assisted living centres or nursing homes;

• older adults. Although older adults are considered 65 years and older, in order to avoid the exclusion of relevant trials, those with a specified inclusion criterion of 60 years of age or over, or if the mean age minus one standard deviation is more than 60 years were included.

Trials were excluded when:

• trial population focused on distinct clinical conditions. Distinct clinical conditions involved any special population with a specific ICD‐10 (10th revision of the International Statistical Classification of Diseases and Related Health Problems) coded diagnosis, such as people with stroke, Parkinson's disease, osteoporosis, or multiple sclerosis. However, because there is a high prevalence of multimorbidity, including dementia, amongst older adults, we included trials looking at multimorbid populations living in the community;

• trial population had moderate (Mini‐Mental State Examination (MMSE) score 11 to 20) or severe (MMSE 0 to 10) dementia (Perneczky 2006). Trials where participants had mild cognitive impairment (MCI) or mild dementia (MMSE 21 to 24) were included. The inclusion of participants with MCI and mild dementia takes into consideration that CBT has been shown effective in that population (Forstmeier 2015).

Types of interventions

Trials were included when they investigated:

• CBT interventions with and without exercise interventions.

We defined CBT interventions as interventions given by a professional or trained non‐professional therapist that were delivered in person, or through telecommunications, such as videoconferencing. The content of the interventions included any therapeutic approaches aimed at modifying maladaptive beliefs and behaviours. We defined CBT interventions based on a summary of six Cochrane Reviews on CBT (Davies 2018; Fuller 2020; Hetrick 2016; James 2020; Simon 2021; Sneddon 2020). This definition is in line with the Prevention of Falls Network Europe (ProFaNE) taxonomy for use in fall prevention (Lamb 2011). CBT interventions were included in this review when they included all the following.

• The goal of the CBT was changing behaviour.

• The CBT component was structured (e.g. there was a protocol to follow).

• The CBT component allowed for interactive discussion.

The CBT intervention was provided by a person trained in following the intervention protocol, not necessarily a certified psychotherapist. The exercise intervention was delivered either alongside at the same time, or separately at a different time by the same or different people, as the CBT component of the combined intervention. Cognitive restructuring may be used in the CBT component to test the assumptions or beliefs about FoF of those undergoing treatment. There may have been a provision of recommended exercises without supervision.

Interventions combining CBT and exercise combined elements of both CBT and exercise interventions for FoF. The exercise intervention component consisted of physical activities that were planned, structured, repetitive, and aimed at improving or maintaining one or more components of physical fitness (CDC 2020; Kendrick 2014). We included sham treatments in the comparison group, defined as interventions that were unlikely to have a substantial impact on physical activity, such as simple lifestyle advice (Kendrick 2014).

Trials were excluded when they:

• did not fulfil our definition of CBT interventions, as described above;

• did not include CBT with and without exercise interventions.

Trials were included when they compared:

• CBT‐only interventions for FoF with standard care, or placebo or sham treatments. Standard care was defined as no change in usual activities (without any additional treatments);

• CBT with exercise interventions for FoF with standard care, or placebo or sham treatments.

Studies with placebo or sham treatments that have active components were included if there was enough information in the study report to determine that these were unlikely to influence behaviour.

Types of outcome measures

Trials assessing FoF and related constructs as primary or secondary outcomes were included in this review. We collected data for the primary outcome at three time points: immediately postintervention, up to and including six months after treatment, and more than six months after treatment; these latter two time points demonstrate sustainability of the effect of the intervention. For secondary outcomes, we only reported immediate postintervention effects.

Primary outcomes

• FoF. As an umbrella term, FoF is documented by different assessment methods focusing on concepts related to emotional (fear) as well as cognitive appraisal (concerns). FoF can be measured through single‐item questionnaires, or through multi‐item measures, such as those for falls efficacy, and balance confidence. Relevant measurement tools to identify FoF are summarised by Kendrick 2014 and Moore 2008. These are:

  • ABC and its associated versions (e.g. its UK version ABC‐UK) (Powell 1995)

  • Falls Efficacy Scale (FES) (Tinetti 1990) and its associated versions (e.g. modified FES (mFES) (Hill 1996), or FES‐I) (Yardley 2005)

  • Geriatric Fear of Falling Measure (GFFM) (Huang 2006)

  • Single‐item Instruments (Tinetti 1990; Howland 1993)

  • Mobility Efficacy Scale (MES) (Lusardi 1997)

  • University of Illinois at Chicago Fear of Falling Measure (UICFFM) (Velozo 2001)

Secondary outcomes

• Activity avoidance (direct questions or questionnaires, such as the Survey of Activities and Fear of Falling in the Elderly (SAFFE) (Lachman 1998) or the Falls Efficacy Scale‐International Avoidance Behaviour (FES‐IAB) (Dorresteijn 2016))

• Occurrence of falls (number of fallers/rate of falls) (Hauer 2006; Lamb 2005)

• Depression (any validated depression‐related questionnaire, e.g. Beck Depression Inventory‐II (BDI‐II)) (Beck 1996)

• Anxiety (any validated anxiety related questionnaire, e.g. Beck Anxiety Inventory (BAI)) (Beck 1988)

• Adverse effects, such as development of new symptoms, distress, or emotional disturbances during intervention (Schermuly‐Haupt 2018)

• Quality of Life (QoL) encompassing related concepts of functional status, health‐related QoL, and well‐being. Examples of relevant measurement tools to identify QoL are:

  • 36‐item Short Form Health Survey (SF‐36) (Ware 1980)

  • World Health Organization Quality of Life‐BREF (WHOQOL‐BREF) (Skevington 2004)

  • Personal Well‐Being Index (PWI) (Cummings 2002)

  • Satisfaction with Life Scale (SWLS) (Diener 1985)

  • Life Satisfaction Questionnaire‐9 (LISAT‐9) (Fugl‐Meyer 1991)

• Burden of treatment (as per the cost of treatment, time spent treating, participant time receiving treatment), and compliance (reported adherence to exercises and presence during psychotherapy)

Search methods for identification of studies

Electronic searches

Using tailored search strategies, we searched the following electronic databases for relevant trials with the main search conducted in March 2021, updated in November 2021 in a second search, and updates on 11 January 2023 in a third search.

• Cochrane Central Register of Controlled Trials (CENTRAL, Issue 1, 2023) via the Cochrane Register of Studies (CRS‐Web)

• MEDLINE Ovid (from 1946 to 11 January 2023)

• Embase Ovid (from 1980 to 11 January 2023)

• CINAHL Plus (Cumulative Index to Nursing and Allied Health Literature) (from 1982 to 11 January 2023)

• PsycINFO (from 1967 to 11 January 2023)

• AMED (Allied and Complementary Medicine from 1985 to 11 January 2023)

There were no limitations based on language or publication status during the initial search. However, only English and German language articles were considered for inclusion, due to the language abilities of the review authors.

In MEDLINE, we combined the subject‐specific terms with the sensitivity‐maximising version of the Cochrane Highly Sensitive Search Strategy for identifying randomised trials (Lefebvre 2019). The search strategies are provided in Appendix 1.

We also searched the following trials registries to identify ongoing and recently completed trials.

• The World Health Organization International Clinical Trials Registry Platform (WHO ICTRP) (11 January 2023)

• ClinicalTrials.gov (11 January 2023)

Searching other resources

We searched reference lists of included articles for relevant articles. We contacted experts in the field for relevant published or unpublished studies. We handsearched the conference proceedings of International Conference on Fall Prevention and Protection (ICFPP) from 2010 to 2022.

Data collection and analysis

Review authors previously or currently involved in conducting a study that was potentially eligible for inclusion in the review were exempt from study selection decisions, risk of bias assessments, and data extraction for their study.

Selection of studies

Two review authors (EL and PU) independently screened titles and abstracts retrieved from the database searches against the inclusion and exclusion criteria using the online screening and data extraction tool Covidence (Covidence). We screened the full texts of the remaining articles. We resolved disagreements for the inclusion or exclusion of titles and abstracts, or study inclusion between the pairs of review authors. When there was no consensus, an independent advisor group (SK, KH, GK, RZ) resolved conflicts. When clarification was required to assess eligibility, we contacted the corresponding authors of the article in question.

Data extraction and management

Two review authors (EL and PU) independently extracted data using an electronic version of a standardised data extraction form for intervention reviews modified to include primary and secondary outcome data. This form was managed in an electronic spreadsheet. The review authors discussed disagreements in data extraction. When there was no consensus, an independent advisor group (SK, KH, GK) resolved conflicts. We piloted the data extraction form on two studies. We considered studies with multiple reports as a single study. We contacted study authors to request missing data.

The standardised data extraction form was based on that of The Cochrane Developmental, Psychosocial and Learning Problems Review Group, as published on their website (Cochrane 2022).

Assessment of risk of bias in included studies

Two review authors (EL, PU) independently assessed the risk of bias of included studies using the Cochrane RoB 1 tool, providing a domain‐based evaluation, as described in Chapter 8 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). The forms were managed in an electronic spreadsheet. We assessed the domains of sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessments, incomplete outcome data, selective outcome reporting, and other potential threats of validity. We contacted corresponding study authors to request any missing information required to classify risk of bias. We resolved disagreements in the assessment of risk of bias between review authors. When there was no consensus, an independent advisor group (SK, KH, GK, RZ) resolved conflicts. Whenever we obtained missing data from study authors, we documented this in the risk of bias table. We considered studies to have an overall high risk of bias when at least three domains were at high risk, or if there was at least one unclear and one high risk in the other domains.

Specifically for trials using cluster randomisation, we considered the additional risk of bias relating to recruitment, baseline imbalance, loss of clusters, incorrect analysis, and comparability with individually randomised trials, as described in Chapter 16 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). Particular attention was given to the randomisation and blinding processes (Eldridge 2016).

Measures of treatment effect

We reported continuous data using the mean difference (MD) with 95% confidence intervals (CI) when studies used the same outcome measures and standardised mean difference (SMD) when studies used different measures for the same clinical outcome. For continuous outcomes, we presented final scores in preference to change scores. Where appropriate, we combined change scores with post‐treatment scores. For dichotomous outcomes, we reported treatment effects as risk ratios (RR), with 95% CIs. We interpreted outcomes presented as SMDs as having small effect sizes for SMD values of 0.2 or greater, medium effect sizes as 0.5 or greater, and large effect sizes as 0.8 or greater in line with the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011).

We considered a one‐month measurement period following the end of the intervention for the postintervention effects due to the logistics of measuring study participants. We chose sustained FoF change time points representing short‐term sustainability (over one month to up to and including six months), and long‐term sustainability (over six months). Sustainability of effects was dichotomised at less than six months and six months or greater as it was considered clinically relevant.

Unit of analysis issues

Unit of analysis issues occur when the number of observations in the analysis do not match the number of units that are randomised, typically occurring in cluster‐randomised trials (Higgins 2019). When studies did not account for clustering in their analysis, an intracluster correlation coefficient (ICC) was estimated of r = 0.02 and the sample size was adjusted accordingly (Killip 2004) for the effective sample size. In taking a conservative approach, we rounded values down, as per Chapter 23 of the Cochrane Handbook for Systematic Reviews of Interventions (Deeks 2022). Studies that presented results over several time periods created unit of analysis error (Higgins 2019). For this reason, we based outcomes in time frames. Caution was taken to avoid analysing count data as dichotomous data, which occurs when outcomes are reported as number of events, such as number of falls, rather than number of participants with these outcomes (Higgins 2019). When studies had multiple experimental arms, and were not a cross‐over design, each arm with a CBT component was given proportional weight, as described in the data synthesis. If cross‐over trials are identified in future updates, they will be handled as described in Chapter 23 of the Cochrane Handbook for Systematic Reviews of Interventions (Deeks 2022).

Dealing with missing data

Missing data were dealt with following the recommendations of the Cochrane Handbook for Systematic Reviews of Interventions (Deeks 2019). We calculated standard deviation (SDs) for continuous outcomes when not reported based on standard errors, CIs, or exact probability (P) values. We contacted study authors to request information where missing data could not be calculated from the information given. We did not impute missing SDs.

Assessment of heterogeneity

We assessed statistical heterogeneity using the Chi² test, considering a low P value (less than 0.1) as statistical evidence of heterogeneity of intervention effects (Deeks 2019). We calculated inconsistencies across studies using the I² statistic, testing the impact of the heterogeneity (Deeks 2019). All calculations used Review Manager 5 (Review Manager 2014). We interpreted results as suggested by Higgins 2019.

• 0% to 40%: heterogeneity might not be important

• 30% to 60%: may represent moderate heterogeneity

• 50% to 90%: may represent substantial heterogeneity

• 75% to 100%: considerable heterogeneity

Comparisons of participant characteristics and methodology of included studies allowed for the evaluation of the clinical and methodological diversity. Visual inspection of forest plots was performed to identify overlapping 95% CIs for consistency of intervention effects.

Assessment of reporting biases

We assessed publication biases using funnel plots where there were more than 10 studies available. Visual inspection for asymmetry was performed to identify the presence of potential bias.

Data synthesis

Analysis of synthesised data followed the recommendations of the Cochrane Handbook for Systematic Reviews of Interventions (Deeks 2022). All calculations were performed using Review Manager 5 using data extracted from the included articles, or obtained directly from study authors (Review Manager 2014).

When considered appropriate, we pooled the results of comparable studies using both fixed‐effect and random‐effects models. We reported random‐effects models as primary analyses, due to their plausible assumptions in medicine, as well as extent of heterogeneity between studies. We included fixed‐effect models in the sensitivity analysis. In studies with multiple arms, we included each arm fulfilling the inclusion criteria (e.g. CBT with and without exercise interventions). We used the number of participants of each of the corresponding intervention arms, divided the number of participants in the control group by the number of included arms of the study. Studies with multiple arms were used throughout. We did not consider pooling data where there was considerable heterogeneity (I² ≥ 75%) that could not be explained by the diversity of methodological or clinical features amongst trials. Where pooling data was inappropriate, we presented trial data in the analyses or tables descriptively and reported these in the text.

We assessed FoF postintervention and to determine sustainability of effect. The postintervention effect is the change in FoF values as measured upon completion of the intervention. Sustainability is "the maintenance of health benefits over time and improving appraisal of existing programs (Shediac‐Rizkallah 1998; Lennox 2020)." We measured sustainability at two time points – up to six months after the intervention and later than six months. We re‐expressed the SMD of the primary outcomes using the FES‐I, calculating the baseline pooled SD of a study whose population was representative of the those included in this review, with the formula from the Cochrane Handbook for Systematic Reviews of Interventions (Deeks 2022).

During the review process, we noted differences in intervention type according to whether exercise was included with CBT. For all outcomes, we presented data separately for CBT with and without exercise interventions in order to better demonstrate the variation between intervention approaches; we also presented a pooled effect estimate for all intervention approaches.

When studies reported several measures for FoF, preference was given for the FES‐I or its derivatives (e.g. modified Falls Efficacy Scale (mFES), FES, Falls Efficacy Scale Swedish variant (FES‐S)). Preference was given to the FES‐I as it was the most used FoF measuring instrument amongst the included studies and is the recommended fall‐related psychological measure for falls efficacy (Moore 2008). This follows the ProFaNE consensus recommendation that psychological consequences of falls should be conceptualised in terms of falls‐related self‐efficacy (Lamb 2005).

Subgroup analysis and investigation of heterogeneity

We selected subgroup characteristics a priori due to their known association with FoF, or influence on the course of intervention outcomes. We performed the following a priori subgroup analyses.

• Age group (aged less than 75 years versus aged 75 years or greater)

• Type of control group (placebo control versus usual care control)

Four a priori subgroup analyses (baseline fall risk, baseline FoF, cognitive state, and sex) were not undertaken for reasons explained in the Differences between protocol and review section.

When there were common trends in interventions that were deemed clinically relevant, we conducted post hoc analyses. Post hoc analyses were:

• CBT only versus CBT with exercise interventions;

• interventions based on AMB versus non‐AMB‐based interventions;

• group versus individual interventions.

Sensitivity analysis

We performed a priori sensitivity analysis for the primary outcome for:

• removing the studies causing considerable heterogeneity (outliers);

• trials with high risks of bias (i.e. with at least three domains at high risk of bias);

• comparing fixed‐effect and random‐effects estimates.

One a priori sensitivity analysis (quasi‐RCTs) was not undertaken for reasons explained in the Differences between protocol and review section.

When there were common trends in studies observed in interventions that were deemed subject to potentially influence quantitative synthesis, we performed post hoc sensitivity analysis for the primary outcome for:

• trials where FoF was not the primary outcome;

• where non‐FES‐based questionnaires were used as measures;

• clustered trials for potential ICC issues;

• trials where active control groups were used.

Summary of findings and assessment of the certainty of the evidence

We prepared Table 1 for CBT interventions (CBT with and without exercise) compared to control for reducing FoF in older people living in the community, listing the outcomes of FoF, immediate postintervention effects; sustainability of effects, up to six months' postintervention; sustainability of effects, more than six months' postintervention; postintervention effects for activity avoidance, occurrence of falls, depression, and adverse effects. We used GRADEpro software to produce the table (GRADEpro GDT).

We used the GRADE approach to assess the certainty of evidence. The certainty of the evidence is rated according to four categories: high, moderate, low, and very low (Schünemann 2019). The decision to define the certainty of evidence is based on five domains for downgrading (risk of bias, inconsistency, indirectness, imprecision, publication bias), with the certainty of evidence initially being rated at high, as all included studies were RCTs (Deeks 2022). A high grade indicates that we are "very confident that the true effect lies close to that of the estimate of the effect;" moderate grade that 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 that "confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect;" and very low that we "have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect" (Schünemann 2019).

Results

Description of studies

Results of the search

We screened 8060 records from the following databases: CENTRAL (1492), MEDLINE (1327), Embase (2007), CINAHL (544), PsycINFO (1168), AMED (504), the WHO International Clinical Trials Registry Platform (415), and ClinicaTtrials.gov (603). Our searches of the reference lists of included studies found one additional study that was not retained for inclusion in the meta‐analysis.

Once duplicates had been removed, we had 4938 records. We excluded 4866 records based on titles and abstracts. We obtained the full text of the remaining 72 records and linked any references pertaining to the same study under a single study ID. Upon further analysis, we excluded 57 studies of which the 24 most potentially relevant excluded studies are described in the Characteristics of excluded studies table. There are three ongoing studies (see Characteristics of ongoing studies table).

We included 12 studies reported in 15 references (Arkkukangas 2019; Dorresteijn 2016; Freiberger 2012; Freiberger 2013; Gitlin 2006; Huang 2011; Parry 2016a; Reinsch 1992; Resnick 2008; Tennstedt 1998; Wetherell 2018; Zijlstra 2009). For a further description of our screening process, see the study flow diagram (Figure 1).

1.

Study flow diagram.

Included studies

Design

Of the 12 included studies, eight studies were individually randomised controlled trials (RCTs), and four were cluster‐RCTs (Freiberger 2013; Reinsch 1992; Resnick 2008; Tennstedt 1998). The study characteristics are summarised in Table 2. There were no quasi‐experimental trials. Most studies were two‐armed RCTs (intervention versus control), except for Arkkukangas 2019 (three‐armed RCT), Huang 2011 (three‐armed RCT), Freiberger 2012 (four‐armed RCT), and Reinsch 1992 (four‐armed RCT). Arkkukangas 2019 had a CBT with exercise arm, an exercise only arm, and a control arm. Huang 2011 had a CBT only arm, a CBT with exercise arm, and a control arm. Freiberger 2012 had two exercise only arms, one CBT with exercise arm, and a control arm. Reinsch 1992 had a CBT alone arm, CBT with exercise arm, exercise alone arm, and control arm. For purposes of this review, we included the CBT only, CBT with exercise, and control arms from the three and four‐armed studies.

1. Summary table, included study characteristics.

Study	Total number	Mean Age	Number of arms	Design	Intervention types	Control type	Home treatment	Total duration and intensity of interventions	Therapist background	Baseline fall rates	FoF outcome measures
Arkkukangas 2019	124	> 75 years	3 (CBT only, CBT with exercise)	RCT	Individual, CBT with exercise	Usual care	Yes	6 hours, bimonthly treatment, for 12 weeks	Physiotherapists	43.2%	FES‐S
Dorresteijn 2016	389	> 70 years	2	RCT	Individual, CBT with exercise (AMB‐based)	Usual care	Yes	15.5 hours, bimonthly treatment, for 16 weeks	Nurses	62.2%	FES‐I
Freiberger 2012	280	> 70 years	4	Cluster‐RCT	Groups of up to 15 people, CBT with exercise (AMB‐based)	Usual care	No	32 hours, weekly treatment, for 16 weeks	Trained "fall prevention instructors"	26.1%	ABC
Freiberger 2013	378	> 65 years	2	RCT	Groups of 5–15 people, CBT with exercise	Usual care	No	16 hours, weekly treatment, for 16 weeks	Physiotherapists and sports scientists	52.3%	FES‐I
Gitlin 2006	319	> 70 years	2	RCT	Individual, CBT with exercise	Usual care	Yes	7.5 hours, monthly treatment, for 24 weeks	Physiotherapists and occupational therapists	—	FES
Huang 2011	186	> 60 years	3	RCT	Groups 8–12 people, CBT with exercise and CBT only	Usual care	No	8–12 hours, weekly treatment, for 8 weeks	Nurses	17.8%	FES
Parry 2016a	415	> 60 years	2	RCT	Individual, CBT only	Usual care	No	6.75 hours, weekly treatment, for 8 weeks	Licenced psychotherapists	—	FES‐I
Reinsch 1992	230	> 60 years	4 (CBT only, CBT with exercise, exercise only)	Cluster‐RCT	Groups 5–25 people, CBT with exercise and CBT only	Social contact and discussion group	No	1 hour, 3 days per week, for 1 year	—	27.1%	Single‐item instruments
Resnick 2008	166	> 60 years	2	Cluster‐RCT	Groups, CBT with exercise	Nutrition education	No	12–18 hours, biweekly treatment, for 12 weeks	Nutritionists or trained laypeople	—	Single‐item instruments
Tennstedt 1998	434	> 60 years	2	Cluster‐RCT	Groups, CBT with exercise	Social contact and discussion group	No	16 hours, biweekly treatment, for 8 weeks	Laypeople	24.9%	mFES
Wetherell 2018	42	> 65 years	2	RCT	Individual, CBT with exercise	Fall prevention education	Yes	8 hours, weekly treatment, for 8 weeks	Physiotherapists with supervision from psychotherapists	64.3%	FES‐I
Zijlstra 2009	540	> 70 years	2	RCT	Groups, CBT with exercise (AMB‐based)	Usual care	no	16 hours, weekly treatment, over 8 weeks	Nurses	55.6%	mFES

ABC: Activities‐Specific Balance Confidence Scale; AMB: "A Matter of Balance"; CBT: cognitive behavioural therapy; FES: Falls Efficacy Scale; FES‐I: FES‐International; FES‐IAB: FES International Avoidance Behaviour; FES‐S: Falls Efficacy Scale Swedish; FoF: fear of falling; mFES: Modified FES; RCT: randomised controlled trial; Single‐Item Instrument: two instruments.

Aim

The primary aim of 10 studies was to investigate the effectiveness of reducing FoF (Arkkukangas 2019; Dorresteijn 2016; Freiberger 2012; Gitlin 2006; Huang 2011; Parry 2016a; Resnick 2008; Tennstedt 1998; Wetherell 2018; Zijlstra 2009). The primary aim of Reinsch 1992 was to investigate the length of time to first fall and the severity of injury associated with the fall; FoF was a secondary outcome. The primary aim of Freiberger 2013 was to reduce falls and injurious falls.

Sample sizes

The 12 included studies involved 3197 participants in both the control groups (total 1544) and intervention groups (total 1653), of whom 2357 (control 1119, intervention 1238) were included for quantitative synthesis, before adjusting for clustering. Study total sample sizes varied from 42 (Wetherell 2018) to 540 participants (Zijlstra 2009).

Setting

The studies were based in six different countries (USA = 5; Germany = 2; the Netherlands = 2; Sweden = 1; Taiwan = 1; UK = 1). One study was an international collaboration between Germany, the USA, and the Netherlands (Freiberger 2012).

Participants

No study noted differences in their baseline sample populations. All studies recruited both sexes. Mean age of recruited participants at baseline varied between 73 (Resnick 2008) and 83 years (Arkkukangas 2019), with one study not reporting the mean age (Huang 2011). The median age of studies included in quantitative synthesis was 77.9 years. The median consisted of all studies except for Huang 2011, who reported baseline age in ranges (60 to 64 years and 65 years or greater) rather than overall mean. Only Resnick 2008 had a mean age of participants of less than 75 years, and only Arkkukangas 2019 had participants with a mean age of more than 80 years. Five studies recruited populations greater than 60 years of mean age (Huang 2011; Parry 2016a; Reinsch 1992; Resnick 2008; Tennstedt 1998), two greater than 65 years (Freiberger 2013; Wetherell 2018), and five greater than 70 years (Arkkukangas 2019; Dorresteijn 2016; Freiberger 2012; Gitlin 2006; Zijlstra 2009). In seven studies, fall experience or the presence of FoF was an inclusion criterion (Dorresteijn 2016; Freiberger 2012; Freiberger 2013; Parry 2016a; Tennstedt 1998; Wetherell 2018; Zijlstra 2009). Inclusion and exclusion criteria of the studies are listed in the Characteristics of included studies table. Studies included community‐dwelling, relatively healthy, and mobile older people without cognitive impairment.

All but three studies reported baseline fall rates (Gitlin 2006; Parry 2016a; Resnick 2008). Most studies stratified participants according to baseline fall history – no falls, one fall, or more than one fall. We dichotomised baseline fall rate as non‐fallers and fallers (fallen once or more) by adding single and multiple fallers. Tennstedt 1998 reported 24.9% of participants with at least one fall in the last three months before baseline. Dorresteijn 2016 had 62.2% and Zijlstra 2009 55.6% of participants falling in the previous six months. Arkkukangas 2019 reported 43.2% and Wetherell 2018 64.3% in the last 12 months. Alternatively, studies stratified by falls versus no falls in past six months (Freiberger 2012: 26.1%) or 12 months (Freiberger 2013: 52.3%; Huang 2011: 17.8%). Reinsch 1992 used the term previous fallers without specifying a timeline (27.1%). The percentage of fallers before the intervention was calculated as the mean between intervention and control groups. The overall weighted mean rate of fallers of all included studies reporting baseline fall rates was 44.4% and the median was 43.2%. Tennstedt 1998 was not included in the calculation for mean and median percentage of fallers in the baseline characteristics, as this study was not included in quantitative synthesis.

Interventions

The interventions fell into two categories: CBT only (Huang 2011; Parry 2016a; Reinsch 1992), and CBT with exercise (Arkkukangas 2019; Dorresteijn 2016; Freiberger 2012; Freiberger 2013; Gitlin 2006; Huang 2011; Reinsch 1992; Resnick 2008; Tennstedt 1998; Wetherell 2018; Zijlstra 2009). Two studies had both a CBT only arm and a CBT with exercise arm that fulfilled the inclusion criteria for inclusion in this review (Huang 2011; Reinsch 1992).

Details of intervention

Three interventions were reported to be directly adapted from or to take elements of the AMB (Dorresteijn 2016; Freiberger 2012; Zijlstra 2009). The interventions focused on using the technique of cognitive restructuring, goal setting, and encouraging the pursuit of activities between sessions in groups while focusing on positive aspects of the exercise. The adaptations to AMB were: Dorresteijn 2016 translated AMB to Dutch and provided a home‐based individual intervention with a motivational interviewing component for establishing an individualised exercise component. The exercise component focused on setting realistic personal goals for increasing activity levels and safe behaviour (using action plans, promoting uptake of old and new activities, and exposure in vivo). Dorresteijn 2016 was a CBT with exercise intervention, as the intervention attempted to plan and structure repetitive purposeful physical activities though focusing on activities that were no longer pursued due to the FoF; Freiberger 2012 used the fall risk education component from AMB and added a cognitive training component; Zijlstra 2009 adapted the intervention to allow for scheduling more time for some activities, changing session frequency from twice to once a week, adding a booster session after six months, and adding more transparencies.

Despite elements overlapping with AMB, other CBT interventions treated FoF differently. Arkkukangas 2019 used interactive motivational interviewing structured using open‐ended questions, affirmations, reflective listening, and summaries (OARS). Freiberger 2013 was adapted for improvement from Freiberger 2012, following the outcomes of their study. Freiberger 2013 developed a fall risk education component that addressed misconceptions about fall risk, attitudes about falls, thoughts, and concerns about falling; negative and positive thinking patterns; identifying potential environmental fall hazards; and dedicated sessions for behavioural changes and attitudes. Gitlin 2006 used the cognitive approach of problem‐solving and reframing. Huang 2011 developed an intervention focusing on restructuring misconceptions to promote a view of fall risk and FoF as controllable through discussions about the associations with falls or FoF. Participant's point of view of FoF (positive and negative aspects about the topic), strategies to manage FoF and family support and implementing it in the participant's daily life, and problem‐solving (during a fall learning how to fall, stand up, and call for help) (Huang 2011). Parry 2016a examined the three Ps model (predisposing, precipitating, and perpetuating) in more detail, formulating the FoF schema. Identified problems were managed using standard non‐specific CBT approaches (e.g. graded exposure for anxiety; activity monitoring, graded activity and behavioural activation for pain, fatigue, and low mood; and sleep management for fatigue) (Parry 2016a).

Reinsch 1992 used a cognitive behavioural approach to improve awareness of environmental hazards and medically related risk factors, through fall prevention education, improve confidence, and lower anxiety, through relaxation training (Reinsch 1992). Resnick 2008 developed the Senior Exercise Self‐efficacy Project (SESEP). This intervention used exposure to enactive mastery experiences (participating in an exercise class), verbal encouragement (counselling and education), and "implementation of interventions to decrease the unpleasant sensations or augment the pleasant sensations associated with exercise" to enhance falls‐related self‐efficacy beliefs and modify outcome expectations. Wetherell 2018 developed the Activity, Balance, Learning, and Exposure (ABLE) intervention. The CBT component was given during the fifth to seventh week and consisted of "psychoeducation about anxiety and the role of avoidance, creation of a fear hierarchy based on identified triggers and avoidance behaviours, exposure practise, cognitive restructuring, and problem‐solving" while using exposure techniques (Wetherell 2018).

Delivery of intervention

Details of interventions are listed in the Characteristics of included studies table. All interventions were guided, and five were individualised treatments (Arkkukangas 2019; Dorresteijn 2016; Gitlin 2006; Parry 2016a; Wetherell 2018), while other studies used group interventions, with mostly between five and 15 participants. Huang 2011 reported up to 12 participants per group, Freiberger 2012 and Freiberger 2013 reported 15 participants, and Reinsch 1992 reported 25 participants. Zijlstra 2009 reported mean group sizes of 10 participants, and Tennstedt 1998 reported 11 participants. Resnick 2008 reported a mean group size of 17 participants (as calculated by 100 participants over six sites). Reinsch 1992 had the greatest variation of group size of between five and 25 participants.

Duration and intensity of interventions

Therapist contact time varied between six hours (Arkkukangas 2019) and 156 hours (Reinsch 1992). Parry 2016a had a total of 6.75 hours; Gitlin 2006 had 7.5 hours; Wetherell 2018 had 8 hours; Huang 2011 had 8 to 12 hours; Dorresteijn 2016 had 15.5 hours; Freiberger 2013, Tennstedt 1998, and Zijlstra 2009 had 16 hours; Resnick 2008 had 12 to 18 hours; and Freiberger 2012 had 32 hours. Intervention intensity varied from three times weekly (Reinsch 1992), twice weekly (Resnick 2008; Tennstedt 1998), weekly (Freiberger 2012; Freiberger 2013; Huang 2011; Parry 2016a; Wetherell 2018; Zijlstra 2009), every two weeks (Arkkukangas 2019; Dorresteijn 2016), and monthly (Gitlin 2006). Interventions were given for 8 weeks (Huang 2011; Parry 2016a; Tennstedt 1998; Zijlstra 2009), 12 weeks (3 months) (Arkkukangas 2019; Resnick 2008), 16 weeks (4 months) (Dorresteijn 2016; Freiberger 2012; Freiberger 2013), 24 weeks (6 months) (Gitlin 2006), and 48 weeks (12 months) (Reinsch 1992). Details of compliance with exercise and psychotherapy components of treatment were lacking amongst the included studies.

Therapist backgrounds

Therapists providing CBT were healthcare professionals of various specialities; physiotherapists (Arkkukangas 2019; Freiberger 2013; Gitlin 2006; Wetherell 2018), nurses (Dorresteijn 2016; Huang 2011; Zijlstra 2009), trained "fall prevention instructors" (Freiberger 2012), sports scientists (Freiberger 2013), occupational therapists (Gitlin 2006), licenced psychotherapists (Parry 2016a), nutritionists (Resnick 2008), or trained laypeople (Resnick 2008; Tennstedt 1998). Wetherell 2018 had a novel approach of providing physiotherapists with supervision from licenced psychotherapists.

Controls

Control groups were mostly usual care (no interventions) (Arkkukangas 2019; Dorresteijn 2016; Freiberger 2012; Freiberger 2013; Gitlin 2006; Huang 2011; Parry 2016a; Zijlstra 2009). Other control groups (placebo controls, considered active control groups) consisted of a non‐FoF‐related discussion group (Reinsch 1992), nutrition education (Resnick 2008), social contact and non‐FoF‐related discussion groups (Tennstedt 1998), and fall prevention education (Wetherell 2018). Fall prevention education was considered a sham treatment, as the education in the context of Wetherell 2018 was designed to be an attention placebo.

Outcomes

All studies reported FoF outcomes, measured using six different instruments; we described these in Table 3. Huang 2011 reported sustainability of effects at two and five months. Freiberger 2012, Gitlin 2006, Tennstedt 1998, and Zijlstra 2009 reported sustainability measures at six months. Dorresteijn 2016, Freiberger 2012, Parry 2016a, Tennstedt 1998, and Zijlstra 2009 reported sustainability measures at 12 months. Arkkukangas 2019 and Freiberger 2012 reported sustainability measures at 24 months. Freiberger 2013 reported 24‐month sustainability measures in a secondary publication. Outcome data were available for most secondary outcomes, except adverse effects and cost of treatment, for which no studies explicitly reported data.

2. Fear of falling measures.

Measure	Studies that used the measure	Score interpretation	Description of measure
Falls Efficacy Scale (FES) (Tinetti 1990)	Gitlin 2006; Huang 2011	Lower scores represent lower concerns about falling (low FoF)	This self‐administered questionnaire is the original FoF scale for which there are many derivates. Items represent the 10 most important activities essential to independent living. This instrument has 10 items, rated on a 1 (not at all concerned) to 10 (very concerned) point Likert scale, with a total range between 10 and 100. Note: Gitlin 2006 added 3 items to this scale from the Activities‐Specific Balance Confidence Scale.
FES‐International (FES‐I) (Yardley 2005)	Dorresteijn 2016; Freiberger 2013; Parry 2016a; Wetherell 2018	Lower scores represent lower concerns about falling (low FoF)	A self‐administered questionnaire in its original version and an expanded, modified version of the FES. It uses 6 additional items representing various levels of challenging and social activities (Moore 2008). The FES‐I instrument has 16 items, rated on a 1 (not at all concerned) to 4 (very concerned) point Likert scale, with a total range between 4 and 64.
Modified FES (mFES) (Hill 1996)	Tennstedt 1998; Zijlstra 2009	Lower scores represent lower concerns about falling (low FoF)	This self‐administered questionnaire is an expanded version of the FES with an additional 4 items representing outdoor activities. This instrument has 14 items, rated on a 1 (not at all concerned) to 4 (very concerned) point Likert scale, with a total range between 4 and 64.
FES Swedish variant (FES‐S) (Hellström 1999; Hellström 2002)	Arkkukangas 2019	Lower scores represent lower concerns about falling (low FoF)	This self‐administered questionnaire is a translated version of the FES with 3 additional items, totalling 13 items rated on a 0 (not confident at all) to 10 (completely confident) point Likert scale, with a total range between 0 and 130.
Activities‐Specific Balance Confidence Scale (ABC) (Powell 1995):	Freiberger 2012	A high score represents a high level of physical functioning (low FoF)	This interviewer‐administered questionnaire has 16 items, rated on a 0% (no confidence) to 100% (completely confident) continuum, with a total score ranging between 0 and 100 (calculated by the total score divided by 16).
Single‐Item Instruments	Reinsch 1992; Resnick 2008	A high score represents high FoF	Reinsch 1992 used a single‐item question rated on a 5‐point scale regarding worry about falling has level of fear ranges from 1 = not at all worried to 5 = extremely worried. Resnick 2008 used a single‐item question asking respondents "How would you rate your fear of falling" on a scale ranging from 0 (no fear) to 4 (very afraid) (Resnick 2008).

FoF: fear of falling.

Only Tennstedt 1998 could not be included in quantitative synthesis, as it did not report sufficient data. They reported mean change scores, with effect sizes missing in the control groups. We contacted the study authors who reported they no longer had access to the data. Freiberger 2012 did not report sufficient data for inclusion; however, when contacted, they provided mean and SD values of the ABC scores so that the study could be included. Wetherell 2018 did not provide sufficient information in their text; however, we could extract the data from their figures, which showed mean and standard errors of the FES‐I scores.

Sources of funding

All studies declared their sources of funding, and no study declared conflicts of interest.

Excluded studies

We excluded 60 articles on inspection of the full text because they did not meet our inclusion criteria (see Figure 1). We reported in the review details of 24 key excluded studies in the Characteristics of excluded studies table. The key excluded studies did not fulfil inclusion criteria due to an ineligible intervention type, as their intervention did not fulfil our definition of CBT (seven studies: Azizan 2015; Brouwer 2003; Dattilo 2014; Gill 2020; Johansson 2018; Lim 2023; van Schooten 2021), or did not have a CBT component (five studies: Headley 2014; Kwon 2011; Lee 2013; Lin 2007; Suttanon 2018); study design (seven studies: Banez 2008; Chen 2014; NCT01268657; Sartor‐Glittenberg 2018; Thiamwong 2019; Walters 2018; Wolfe 2018); and control group (five studies: Duenas 2019; IRCT20211201053248N1; Liu 2014; NCT02727374; NCT03211429).

Studies awaiting classification

There are no studies awaiting classification.

Ongoing studies

Three studies are likely to fulfil inclusion criteria but were not included in this review as they are still ongoing.

The Own your Balance Study is a three‐armed RCT investigating eHealth options to provide a CBT with a tailored exercise programme (using components from the StandingTall Balance Confidence study) for reducing FoF to community living older adults (ACTRN12621000440820). This study recruited 189 participants and is expected to finish collecting data by July 2023 (www.neura.edu.au/project/own-your-balance/).

NCT05192408 is a multicomponent intervention RCT for reducing FoF in community‐dwelling older adults that began recruitment in January 2022. The primary outcome of this trial is FoF measured using the short FES‐I. The secondary outcomes are number of falls and physical activity measured using the Incidental and PlannEd activity Questionnaire – Weekly Average (IPEQ‐WA). The study plans on recruiting 420 participants, and is expected to be finished in May 2023.

Taylor 2021 is comparing motivational interviewing (MI) with standard care in community‐dwelling older adults after hip fracture (MIHip trial). The primary outcome of this trial is changes in daily time spent walking; however, one of several secondary outcomes is FoF measured using the FES. Although the intervention is MI, it is considered to be a CBT intervention, as its goal is to change behaviour, there is a protocol to follow, it is designed to allow for interactive discussion, and delivered by trained therapists. The trial recruited around 270 participants. Data collection is expected to be completed by 2023.

Risk of bias in included studies

The findings of the risk of bias assessment are presented in Figure 2 (risk of bias summary) and Figure 3 (risk of bias graph). Detailed findings are reported in the Characteristics of included studies table.

2.

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

3.

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

Allocation

Random sequence generation

All studies reported random allocation of participants to the intervention and control groups. We judged the risk of bias in random sequence generation to be low in 10/12 studies as the method of random sequence generation was clearly described. It was unclear in two studies as the exact method was not described (Reinsch 1992; Tennstedt 1998). No study had a high‐risk rating.

Allocation concealment

Nine studies had a low risk of allocation concealment, showing that participants and investigators could not foresee assignments. One study was at high risk of bias for allocation concealment as the cluster randomisation procedure was conducted prior to recruitment and inclusion of participants, leading to identification/recruitment bias (Freiberger 2013). Two studies with cluster randomisation were at unclear risk of bias as the timing of the randomisation and the inclusion of participants was not specified (Reinsch 1992; Tennstedt 1998).

Blinding

Performance bias

For psychological interventions, it is difficult to blind participants and therapists to the intervention being provided. None of the trials reported blinding of participants or personnel providing therapies. Therefore, for the domain of performance bias, we considered all 12 studies at high risk of bias.

Detection bias

All outcome measurement instruments were appropriate. The assessment of FoF and the secondary outcomes of depression, anxiety, and QoL tend to be assessed from the subjective participant's perspective, and rely on self‐reports by participants who know their group allocation, which leads to a judgement of some concerns. Measures were self‐reported, and as such impossible to blind. However, eight trials minimised the likelihood of observer bias by blinding the outcome assessors, during data analysis, to group allocation, and no study reported unsuccessful blinding of assessors (Arkkukangas 2019; Dorresteijn 2016; Freiberger 2012; Gitlin 2006; Huang 2011; Resnick 2008; Wetherell 2018; Zijlstra 2009). Hence, these trials were judged at unclear risk of bias. Tennstedt 1998 was at unclear risk of bias as they blinded the assessors, but outcomes were self‐reported by participants who knew their group allocation. For two trials, authors reported that outcome assessors were not blinded to group allocation (Freiberger 2013; Parry 2016a), and Reinsch 1992 did not report whether the outcome assessor was blinded, and did not report specific statistical analysis considering cluster‐randomisation, so we classified these studies at high risk of bias for this domain.

Incomplete outcome data

Nine trials were at low risk of attrition bias due to low or balanced dropout rates or adequately described intention‐to‐treat analyses with missing data imputation. Three studies were at high risk of bias due to high dropout rates and no missing data imputation (Parry 2016a; Reinsch 1992; Resnick 2008).

Selective reporting

Six trials reported all prespecified outcomes in the trial registration or protocol, so we judged them at low risk of reporting bias (Arkkukangas 2019; Dorresteijn 2016; Freiberger 2012; Freiberger 2013; Parry 2016a; Zijlstra 2009). We judged the risk of bias to be unclear in the other six trials because the study protocol was not published, and verification of preplanned analyses was not possible (Gitlin 2006; Huang 2011; Reinsch 1992; Resnick 2008; Tennstedt 1998; Wetherell 2018).

Other potential sources of bias

Nine studies had low risk of other bias. Three studies were at unclear risk of other bias (Reinsch 1992; Resnick 2008; Wetherell 2018). Reinsch 1992 had a disproportionate ratio of females to males in the control group compared to the intervention groups and did not adjust analyses for clustering. Resnick 2008 also did not adjust for clustering. Wetherell 2018 only partly reported their outcomes in figures within the manuscript. As this reduces the usability of the published data and introduces a potential margin of error by having to extract the data from figures, this study was rated at unclear risk of other sources of bias. One study had a high risk of other bias due to incomplete outcome data and non‐transparent reporting of intention‐to‐treat analyses and analyses for compliant subgroups (Tennstedt 1998). Additionally, Tennstedt 1998 did not adjust for clustering. This lack of reported outcomes prevented this study from being included in quantitative synthesis.

Effects of interventions

See: Table 1

See Table 1 for the main comparison: CBT interventions for reducing fear of falling in community‐dwelling older people.

Primary outcome

Fear of falling

We pooled data from 11/12 included studies (all but Tennstedt 1998) (2357 participants; 1238 in intervention arm, 1119 in control arm) in a meta‐analysis for CBT interventions for reducing FoF. There were clustering issues in three studies (Freiberger 2013; Reinsch 1992; Resnick 2008). Freiberger 2013 incorporated an analysis design using a three‐level linear mixed‐effects model. Their reported data adequately adjusted for clustering and were incorporated directly in the quantitative synthesis of this review. Reinsch 1992 did not address the cluster design; therefore, we adjusted their data: for the FoF outcomes of the CBT only group and CBT with exercise group, we adjusted the denominator values for participants in the intervention groups from 32 to 28 participants and from 50 to 43 participants, respectively. We similarly adjusted the control group denominator values from 23 to 20 and then divided the control group when comparing with each of the interventions group within the same analysis (CBT only and CBT with exercise; 10 participants for each subgroup). Resnick 2008 did not adjust for clustering, as they considered that the clustering resulted in significant outcome differences. However, for purposes of this review, we adjusted their sample sizes to account for the clustering effect. We adjusted the intervention and control groups from 64 to 56 participants and from 39 to 34 participants, respectively.

Immediate postintervention effects on fear of falling

When pooled using the SMD, we found that CBT with and without exercise interventions versus control probably reduces FoF (SMD −0.23, 95% CI −0.36 to −0.11; I² = 48%; 11 studies, 2357 participants; moderate‐certainty evidence; Analysis 1.1). Two studies had a CBT‐only arm and a CBT with exercise arm (Huang 2011; Reinsch 1992). We re‐expressed the SMD on the FES‐I using an SD of 9.41 from the baseline measures of Parry 2016a. This corresponds to a reduction on the FES‐I scale of 2.2 (95% CI −3.4 to −1.0). Using a rule of thumb interpretation, we judged the size of this effect to be small. We judged this evidence to be moderate certainty, downgrading one level for serious risk of bias across all studies due to performance and detection bias; and due to the nature of psychological interventions, studies included unavoidable high risk of performance and detection bias.

1.1. Analysis.

Comparison 1: Fear of falling: immediate postintervention, Outcome 1: Fear of falling: subgrouped according to intervention approach

CBT‐only interventions probably reduce FoF (SMD −0.31, 95% CI −0.56 to −0.05; I² = 25%; 3 studies, 472 participants; Analysis 1.1); this included measures from the FES‐I, FES, and single‐item instrument. We found a similar result with reduced FoF when CBT interventions included an exercise component (SMD −0.22, 95% CI −0.36 to −0.07; I² = 53%; 10 studies, 1885 participants; Analysis 1.1); assessments included measures from FES‐S, mFES, ABC, FES‐I, FES, and single‐item instruments.

CBT‐only interventions showed a low level of heterogeneity (I² less than 40%), the CBT with exercise interventions and the pooled effect sizes of both interventions show a moderate heterogeneity (I² greater than 40%). Formal tests of subgroup differences according to whether exercise was included in the intervention indicated no differences between the two intervention types (P = 0.53). We conducted all prespecified sensitivity analyses (Analysis 1.1). Whilst statistical heterogeneity was sometimes reduced, the effect sizes were similar to our primary analysis, and we did not alter our interpretation of the effect as a result of sensitivity analysis (see Table 4).

3. Summary table, sensitivity analysis.

Analysis	Outcomes (SMD)
Immediate postintervention effects
Removing the studies causing considerable heterogeneity (outliers)	−0.30 (95% CI −0.39 to −0.22; P = 0.65, I2 = 0%; 12 studies, 2228 participants)
Trials with high risks of bias (i.e. ≥ 3 domains of high risk of bias)	−0.17 (95% CI −0.32 to −0.01; P = 0.04, I2 = 50%; 9 studies, 1625 participants)
Comparing fixed‐effect and random‐effects estimates	−0.26 (95% CI −0.35 to −0.18; P = 0.03, I2 = 48%; 13 studies, 2357 participants)
Trials where FoF was not the primary outcome	−0.23 (95% CI −0.37 to −0.10; P = 0.01, I2 = 58%; 12 studies, 2266 participants)
Where non‐FES‐based questionnaires were used as measures	−0.31 (95% CI −0.40 to −0.20; P = 0.48, I2 = 0%; 9 studies, 2047 participants)
Clustered studies for potential ICC issues	−0.31 (95% CI −0.40 to −0.20; P = 0.48, I2 = 0%; 9 studies, 2047 participants)
Trials where active control groups were used	−0.25 (95% CI −0.40 to −0.10; P = 0.006, I2 = 63%; 9 studies, 2136 participants)
Sustainability of effects, up to 6 months' postintervention
Removing the studies causing considerable heterogeneity (outliers)	−0.31 (95% CI −0.44 to −0.17; P = 0.14, I2 = 36%; 8 studies, 1650 participants)
Trials with high risks of bias (i.e. ≥ 3 domains of high risk of bias)	−0.23 (95% CI −0.43 to −0.03; P = 0.003, I2 = 68%; 8 studies, 1470 participants)
Comparing fixed‐effect and random‐effects estimates	−0.27 (95% CI −0.36 to −0.17; P = 0.005, I2 = 63%; 9 studies, 1784 participants)
Trials where active control groups were used	−0.23 (95% CI −0.40 to −0.05; P = 0.004, I2 = 67%; 8 studies, 1744 participants)
Sustainability of effects, more than 6 months' postintervention
Removing the studies causing considerable heterogeneity (outliers)	−0.27 (95% CI −0.42 to −0.12; P = 0.20, I2 = 36%; 4 studies, 1117 participants)
Trials with high risks of bias (i.e. ≥ 3 domains of high risk of bias)	−0.24 (95% CI −0.39 to −0.08; P = 0.3, I2 = 19%; 4 studies, 871 participants)
Comparing fixed‐effect and random‐effects estimates	−0.28 (95% CI −0.40 to −0.17; P = 0.32, I2 = 14%; 5 studies, 1185 participants)

AMB: "A Matter of Balance"; CI: confidence interval; SMD: standardised mean difference.

Subgroup analyses

In formal tests for subgroup interactions, we found no differences in both a priori or post‐hoc analysis findings according to the mean age of participants (note: Huang 2011 was excluded, as they did not report mean age of participants) (Analysis 1.2), the type of control (Analysis 1.3), if their interventions were based on AMB or not (Analysis 1.4), or individual or group interventions (Analysis 1.5); the 95% CI in the subgroups clearly overlapped, which was reflected by large P values for the test for subgroup differences. These results remained true when the outlier, Freiberger 2012, was removed (see Table 5).

1.2. Analysis.

Comparison 1: Fear of falling: immediate postintervention, Outcome 2: Fear of falling: subgrouped according to mean age

1.3. Analysis.

Comparison 1: Fear of falling: immediate postintervention, Outcome 3: Fear of falling: subgrouped according to control

1.4. Analysis.

Comparison 1: Fear of falling: immediate postintervention, Outcome 4: Fear of falling: subgrouped according to "A Matter of Balance" (AMB) versus non‐AMB based interventions

1.5. Analysis.

Comparison 1: Fear of falling: immediate postintervention, Outcome 5: Fear of falling: subgrouped according to group versus individual interventions

4. Summary table, subgroup analysis.

Analysis	Subgroup	Outcomes (SMD)
Immediate postintervention effects
Age group	< 75 years	−0.11 (95% CI −0.54 to −0.32; 1 study, 90 participants)
	≥ 75 years	−0.31 (95% CI −0.40 to −0.22; P = 0.61, I2 = 0%; 9 studies, 1962 participants)
Type of control group	Placebo control	−0.32 (95% CI −0.41 to −0.23; P = 0.45, I2 = 0%; 8 studies, 2007 participants)
	Usual care	−0.13 (95% CI −0.42 to −0.16; P = 0.94, I2 = 0%; 4 studies, 221 participants)
Interventions based on AMB	AMB‐based studies	−0.31 (95% CI −0.45 to −0.17; P = 0.44, I2 = 0%; 2 studies, 783 participants)
	Non‐AMB studies	−0.30 (95% CI −0.41 to −0.19; P = 0.53, I2 = 0%; 10 studies, 1445 participants)
Group vs individual interventions	Group	−0.30 (95% CI −0.42 to −0.18; P = 0.65, I2 = 0%; 7 studies, 1125 participants)
	Individual	−0.30 (95% CI −0.43 to −0.18; P = 0.34, I2 = 11%; 5 studies, 1103 participants)
Sustainability of effects, up to 6 months' postintervention
Type of control group	Placebo control	−0.55 (95% CI −1.18 to −0.09; 1 study, 40 participants)
	Usual care	−0.30 (95% CI −0.43 to −0.16; P = 0.11, I2 = 42%; 7 studies, 1610 participants)
Interventions based on AMB	AMB‐based studies	−0.39 (95% CI −0.58 to −0.13; P = 0.68, I2 = 0%; 2 studies, 734 participants)
	Non‐AMB studies	−0.26 (95% CI −0.46 to −0.07; P = 0.11, I2 = 44%; 6 studies, 916 participants)
Group vs individual interventions	Group	−0.42 (95% CI −0.68 to −0.17; P = 0.19, I2 = 40%; 4 studies, 598 participants)
	Individual	−0.25 (95% CI −0.39 to −0.10; P = 0.27, I2 = 23%; 5 studies, 1052 participants)
Sustainability of effects, more than 6 months' postintervention
Interventions based on AMB	AMB‐based studies	−0.29 (95% CI −0.43 to −0.14; P = 0.45, I2 = 0%; 2 studies, 717 participants)
	Non‐AMB studies	−0.16 (95% CI −0.64 to 0.31; P = 0.04, I2 = 76%; 2 studies, 400 participants)
Group vs individual interventions	Group	−0.24 (95% CI −0.43 to −0.04; 1 study, 405 participants)
	Individual	−0.27 (95% CI −0.50 to −0.03; P = 0.11, I2 = 54%; 3 studies, 712 participants)

AMB: "A Matter of Balance".

Publication bias

There was no significant publication bias suspected. Visual inspection of the funnel plot of the postintervention effects of CBT interventions for FoF showed no asymmetry. This was considered valid as there were over 10 studies, and study populations varied (see sample sizes) (Deeks 2022).

Sustainability of fear of falling effects up to six months' postintervention

Eight of 11 studies included in quantitative synthesis reported sustainability of effects, up to six months' postintervention (Arkkukangas 2019; Dorresteijn 2016; Freiberger 2012; Gitlin 2006; Huang 2011; Parry 2016a; Wetherell 2018; Zijlstra 2009). One study had two intervention arms (Huang 2011). When pooled using the SMD, we found that CBT with and without exercise interventions versus control may sustain a reduction FoF up to six months' postintervention (SMD −0.24, 95% CI −0.41 to −0.07; I² = 63%; 8 studies, 1784 participants; low‐certainty evidence; Analysis 2.1). We re‐expressed the SMD on the FES‐I, using an SD of 9.41 from the baseline measures of Parry 2016a. This corresponds to a reduction on the FES‐I scale of 2.3 (95% CI −4.3 to −0.19). Using a rule of thumb interpretation, we judged the size of this effect to be small. We judged this evidence to be of low certainty, downgrading by one level owing to serious risk of bias across all studies due to performance and detection bias; and due to inconsistency in the CBT with exercise subgroup having substantial heterogeneity.

2.1. Analysis.

Comparison 2: Fear of falling: sustainability of effects, up to 6 months' postintervention, Outcome 1: Fear of falling: subgrouped according to intervention approach

The sustainability of effects of CBT‐only interventions versus control were assessed using the FES and FES‐I. We found that FoF may be reduced (SMD −0.27, 95% CI −0.47 to −0.07; I² = 0%; 2 studies, 404 participants; Analysis 2.1). The sustainability of effects of CBT with exercise interventions were assessed using the ABC, FES‐S, mFES, FES‐I, and FES. We found that FoF may be reduced (SMD −0.24, 95% CI −0.46 to −0.02; I² = 72%; 7 studies, 1380 participants; Analysis 2.1).

There was no heterogeneity with CBT‐only interventions (0%), with the CBT with exercise interventions alone and the pooled effect sizes of both interventions showed moderate heterogeneity (I² greater than 50%). Formal tests of subgroup differences according to whether exercise was included in the intervention indicated no differences between the two intervention types (P = 0.84). We conducted the prespecified sensitivity analyses (Analysis 2.1). Whilst statistical heterogeneity was sometimes reduced, the effect sizes were similar to our primary analysis, and we did not alter our interpretation of the effect as a result of sensitivity analysis (see Table 4).

Subgroup analyses

In formal tests for subgroup interactions, we found no differences in a priori or post‐hoc analyses findings according to the type of control (Analysis 2.2), if their interventions were based on AMB or not (Analysis 2.3), or individual or group interventions (Analysis 2.4); the 95% CI in the subgroups clearly overlapped, which was reflected by large P values for the test for subgroup differences. This remains true when the outlier, Freiberger 2012, was removed (see Table 5). Subgroup analysis could not be performed for mean age of study participants, as all studies reporting on sustainability over six months had a baseline mean age of 75 years and over.

2.2. Analysis.

Comparison 2: Fear of falling: sustainability of effects, up to 6 months' postintervention, Outcome 2: Fear of falling: subgrouped according to control

2.3. Analysis.

Comparison 2: Fear of falling: sustainability of effects, up to 6 months' postintervention, Outcome 3: Fear of falling: subgrouped according to AMB versus non‐AMB based interventions

2.4. Analysis.

Comparison 2: Fear of falling: sustainability of effects, up to 6 months' postintervention, Outcome 4: Fear of falling: subgrouped according to group versus individual interventions

Publication bias

As with the outcomes of immediate postintervention effects, there was no significant publication bias suspected of the outcomes of sustainability of effects, up to six months' postintervention. Visual inspection of the funnel plot of the postintervention effects of CBT interventions for FoF showed no asymmetry. However, in containing nine studies, this analysis may lack validity (Deeks 2022).

Sustainability of fear of falling effects more than six months' postintervention

Five of 11 studies included in quantitative synthesis reported sustainability of effects up to six months' postintervention (Arkkukangas 2019; Dorresteijn 2016; Freiberger 2012; Parry 2016a; Zijlstra 2009). When pooled using the SMD, we found that CBT with and without exercise interventions versus control probably sustains a reduction in FoF more than six months'' postintervention (SMD −0.28, 95% CI −0.40 to −0.15; I² = 14%; 5 studies, 1185 participants; moderate‐certainty evidence; Analysis 3.1). We re‐expressed the SMD on the FES‐I, using an SD of 9.41 from the baseline measures of Parry 2016a. This corresponds to a reduction on the FES‐I scale of 2.6 (95% CI −3.8 to −1.8). Using a rule of thumb interpretation, we judged the size of this effect to be small. We judged this evidence to be moderate certainty downgrading one level owing to serious risk of bias across all studies due to performance and detection bias; due to the nature of psychological interventions, studies included unavoidable high risk of performance and detection bias.

3.1. Analysis.

Comparison 3: Fear of falling: sustainability of effects, more than 6 months' postintervention, Outcome 1: Fear of falling: subgrouped according to intervention approach

The sustainability of effects of CBT‐only interventions versus control were assessed using the FES‐I. We found that CBT‐only interventions probably reduces FoF (SMD −0.38, 95% CI −0.60 to −0.15; 1 study, 314 participants; Analysis 3.1). The sustainability of effects of CBT with exercise interventions versus control were assessed used the FES‐I, FES‐S, and mFES. We found that CBT with exercise interventions probably reduces FoF (SMD −0.24, 95% CI −0.39 to −0.08; I² = 19%; 4 studies, 871 participants; Analysis 3.1).

Formal tests of subgroup differences according to whether exercise was included in the intervention indicated no differences between the two intervention types (P = 0.32). We conducted the prespecified sensitivity analyses on Analysis 3.1. Whilst statistical heterogeneity was sometimes reduced, the effect sizes were similar to our primary analysis, and we did not alter our interpretation of the effect as a result of sensitivity analysis (see Table 4).

Subgroup analyses

In formal tests for subgroup interactions, we found no differences in both a priori or post‐hoc analysis findings according to if their interventions were based on AMB or not (Analysis 3.2), or individual or group interventions (Analysis 3.3). The 95% CIs in the subgroups clearly overlapped, which was reflected by large P values for the test for subgroup differences. This remained true when the outlier, Freiberger 2012, was removed (see Table 5). Subgroup analysis for type of control group and mean age could not be conducted as all studies reporting sustainability over six months used control groups with usual care, and had a baseline mean age of study participants of 75 years and over.

3.2. Analysis.

Comparison 3: Fear of falling: sustainability of effects, more than 6 months' postintervention, Outcome 2: Fear of falling: subgrouped according to "A Matter of Balance" (AMB) versus non‐AMB based interventions

3.3. Analysis.

Comparison 3: Fear of falling: sustainability of effects, more than 6 months' postintervention, Outcome 3: Fear of falling: subgrouped according to group versus individual interventions

Publication bias

As with the outcomes of immediate postintervention effects and sustainability of effects up to six months' postintervention, there was no significant publication bias suspected of the outcomes of sustainability of effects more than six months' postintervention. Visual inspection of the funnel plot of the postintervention effects of CBT interventions for FoF did not show asymmetry. However, in containing five studies, this analysis may lack validity (Deeks 2022).

Secondary outcomes

Activity avoidance

Two studies reported on activity avoidance; however, due to missing information in Tennstedt 1998, only Dorresteijn 2016 could be included in quantitative synthesis. Tennstedt 1998 reported the Sickness Impact Profile (SIP), a 68‐item scale measuring the changes in a person's behaviour because of health problems. Dorresteijn 2016 reported on activity avoidance measured using the FES‐IAB. CBT‐only interventions may reduce the level of activity avoidance (MD −2.57, 95% CI −4.67 to −0.47; 1 study, 312 participants; low‐certainty evidence; Analysis 4.1). We downgraded the evidence one level owing to serious risk of bias due to performance and detection bias, and due to imprecision, as there were fewer than 200 participants in each study group (Cochrane 2022).

4.1. Analysis.

Comparison 4: Secondary outcomes, Outcome 1: Activity avoidance: immediate postintervention

Occurrence of falls

Eight studies reported the occurrence of falls (Arkkukangas 2019; Dorresteijn 2016; Huang 2011; Parry 2016a; Reinsch 1992; Tennstedt 1998; Wetherell 2018; Zijlstra 2009); however, Parry 2016a, Reinsch 1992, and Tennstedt 1998 did not publish sufficient data to be included in the quantitative synthesis. Huang 2011 published outcomes from two intervention arms (CBT only and CBT with exercise) and, as such, appeared twice in the analysis. The outcomes were dichotomised (falls versus no falls) and assessed using a random‐effects model, calculating an RR with 95% CIs. The CBT‐only intervention had an RR of 0.50 (95% CI 0.21 to 1.20; 1 study, 90 participants). The CBT with exercise interventions had an RR of 1.04 (95% CI 0.70 to 1.53; I² = 82%; 5 studies, 1029 participants). The overall pooled RR of CBT with and without exercise interventions was 0.96 (95% CI 0.66 to 1.39; I² = 80; 5 studies (of which Huang 2011 had two intervention arms), 1119 participants; very low‐certainty evidence; Analysis 4.2). We are uncertain whether CBT with and without exercise interventions reduces falls. We downgraded the evidence one level owing to serious risk of bias across all studies due to performance and detection bias; one level due to serious inconsistency owing to heterogeneity existing between studies in the CBT with exercise groups; and one level for serious imprecision owing to CI indicating both benefit and harm.

4.2. Analysis.

Comparison 4: Secondary outcomes, Outcome 2: Occurrence of falls: immediate postintervention

The pooled RR is inconclusive, not ruling out either benefit or harm of the CBT interventions for reducing the risk of falling.

Depression

Two studies reported depression using the HADS‐D subscore for depression (Parry 2016a) and GDS (Resnick 2008) (Analysis 4.3). Parry 2016a reported that their CBT‐only intervention may reduce depression (MD −1.26, 95% CI −1.96 to −0.56; 1 study, 314 participants; Table 6). The mean HADS‐D control group score was 0.1 (SD 2.65). Resnick 2008 did not adjust for clustering for depression outcomes. For the purpose of this review, we adjusted their data for clustering, adjusting the intervention and control groups to 56 and 34 from 64 and 39 participants, respectively. Resnick 2008 reported that their CBT with exercise intervention may also reduce depression (MD −0.38, 95% CI −0.80 to 0.04; 1 study, 90 participants; Table 6). The average GDS control group score was 0.79 (SD 1.1). As there was only one study in each intervention subgroup reporting depression outcomes, we reported these as separate analyses in Table 1. When we pooled the results from the two studies using an SMD, we found that CBT with and without exercise may reduce the level of depression (SMD −0.41, 95% −0.60 to −0.21; 2 studies, 404 participants; low‐certainty evidence; Analysis 4.3). Using a rule of thumb interpretation, we judged the size of this effect to be small. We downgraded the evidence one level owing to serious risk of bias across both studies due to performance and detection bias; and one level for serious imprecision owing to having fewer than 200 participants (combined) in the control arm of the included studies.

4.3. Analysis.

Comparison 4: Secondary outcomes, Outcome 3: Depression: immediate postintervention

5. Mean difference outcomes, immediate postintervention effects for depression.

Study	Intervention mean	Intervention SD	Intervention total	Control mean	Control SD	Control total	Mean difference IV, random, 95% CI
Parry 2016a	−1.16	3.54	151	0.1	2.65	163	−1.26 (95% CI −1.96 to −0.56)
Resnick 2008	0.41	0.79	56	0.79	1.1	34	−0.38 (95% CI −0.80 to 0.04)

CI: confidence interval; SD: standard deviation.

Anxiety

Only one study reported anxiety measured using the HADS subscore for anxiety (Parry 2016a). The effect size was not conclusive (MD 0.06, 95% CI −0.60 to 0.72; 1 study, 344 participants; Analysis 4.4).

4.4. Analysis.

Comparison 4: Secondary outcomes, Outcome 4: Anxiety: immediate postintervention

Adverse effects

No studies measured adverse effects associated with their interventions.

Quality of life

Four studies measured QoL using various measures: WHOQOL‐BREF (26‐item scale, higher score represents higher quality of life) (Huang 2011), WHOQOL‐Older Adults Module (WHOQOL‐OLD, 24‐item scale, higher score represents higher quality of life) (Parry 2016a), self‐rated health questionnaire (1‐item scale, rated 1 to 5, lower score represent greater health) (Reinsch 1992), and the 12‐item Short Form Health Survey (SF‐12; higher score indicate better functioning) (Resnick 2008). Parry 2016a used three measures of quality of life; WHOQOL‐OLD, the Euro quality of life‐5 Dimension (EQ‐5D (5L)), and the Short‐Form Six‐Dimension (SF‐6D). The WHOQOL‐OLD was given preference over the EQ‐OLD, EQ‐5D, and SF‐6D due to its specific focus on aged populations (Siette 2021). The SF‐12 used by Resnick 2008 consisted of physical well‐being and mental health well‐being mean (SD) scores. The physical well‐being and mental health well‐being mean (SD) subscores were combined for data synthesis.

CBT‐only interventions had an inconclusive effect size (SMD −0.04, 95% CI −0.23 to 0.14; I² = 0%; 3 studies, 472 participants; Analysis 4.5). CBT with exercise interventions also had an inconclusive effect size (SMD −0.47, 95% CI −0.98 to 0.05; I² = 67%; 3 studies, 229 participants; Analysis 4.5). Their pooled effect size was inconclusive (SMD −0.24, 95% CI −0.51 to 0.04; I² = 58%; 4 studies (of which two had a CBT‐only arm and CBT with exercise arm), 701 participants; Analysis 4.5).

4.5. Analysis.

Comparison 4: Secondary outcomes, Outcome 5: Quality of life: immediate postintervention

We are uncertain whether CBT with and without exercise interventions improves QoL. We judged this evidence to be of very low certainty, downgrading one level owing to serious risk of bias across all studies due to performance and detection bias; one level due to serious inconsistency owing to heterogeneity existing between studies in the CBT with exercise groups; and one level for serious imprecision owing to CIs being compatible with either benefit or harm of the intervention.

Burden of treatment

No studies reported data to assess the burden of treatment.

Discussion

Summary of main results

The included studies assessed the effects of CBT in reducing FoF in older people living in the community. This review identified 12 studies, of which 11 were included for quantitative synthesis, totalling 2383 participants after adjustment for clustering (1250 in intervention arm, 1133 in control arm instead of 1273 in intervention arm and 1146 in control arm), that investigated the effects of CBT with and without exercise interventions versus control. A summary of the evidence is presented in Table 1.

The immediate postintervention effect, and sustainability of effect up to six months and more than six months showed a small reduction in effect size. The certainty of evidence was moderate (immediate effect), low (short‐term sustainability), and moderate (long‐term sustainability). These results suggest that CBT with and without exercise interventions probably reduces FoF following the end of treatment and over six months, compared to control. The benefits up to six months may be sustained following cessation of the intervention. Sensitivity analyses show that the conclusions from the primary analysis remain stable.

Immediate postintervention effects and sustainability of effects up to six months showed moderate levels of heterogeneity. Considering the variations in intervention designs (CBT only, CBT with exercise, duration, and group or individual interventions) and population characteristics (in particular age and proportion of participants falling at baseline), the clinical diversity may be a significant source of heterogeneity.

Heterogeneity could not be explained through subgroup analyses separating studies based on age group, baseline fall rate, type of control group, AMB and non‐AMB interventions, individual and group interventions, and duration of intervention through formal tests for subgroup interactions.

We conducted secondary analyses for activity avoidance, occurrence of falls, depression, anxiety, adverse effects, and QoL. Although only one study reported activity avoidance, it found that CBT‐only interventions for FoF may reduce activity avoidance. However, we are uncertain if the occurrence of falls, depression, anxiety, or QoL are changed following intervention, due to very low‐certainty evidence. No study reported information about adverse effects (harms).

Overall completeness and applicability of evidence

The studies included in this review are likely most representative of high‐income countries. However, this review is unlikely to be internationally representative, due to a lack of research conducted on populations in low‐ to middle‐income countries, despite studies showing a high prevalence of FoF in low‐ to middle‐income countries (Birhanie 2021). This is significant, as FoF is higher in minority ethnic groups, lower educational groups, and lower household income groups (Kumar 2014). These groups should receive particular attention when implementing such treatments in clinical practice. The included studies were heterogeneous, showing a variation in study population, such as mean age and baseline fall rates, demonstrating possible variations in recruitment approaches, potentially reducing the true generality of community‐dwelling older adults.

Tennstedt 1998, which used a CBT with exercise intervention called A Matter of Balance (AMB), was a large study with 434 participants that could not be included in quantitative synthesis due to insufficient data published in their manuscript. The study published FoF effect size, as measured by the FES, for only the intervention group, and not the control group. We were informed by study authors that the data were likely destroyed. This study showed positive results of their intervention. The omission of this study likely reduces the certainty of the results presented in this review. This is additionally significant, as three other studies included in this review were based on, or incorporated elements of, AMB (Dorresteijn 2016; Freiberger 2012; Zijlstra 2009).

Detailed protocols of the different interventions included in this review were generally not publicly available, and as such limit the qualitative analysis of the descriptions of the studies. Therefore, the individual intervention's components were difficult to compare, particularly when all but Parry 2016a were delivered by trained health professionals who were not registered psychotherapists and must therefore have followed a treatment protocol. The lack of greater detail of the interventions limits how they can be compared, concerning, for example, duration of CBT and exercise components.

The minimal clinically important difference (MCID) has not yet been established for FoF measures in the populations used in studies included in this review. The MCID is a measure to demonstrate genuine clinical improvement, rather than statistical improvement (Bloom 2023). However, this value does have limitations, as it can vary significantly based on how it is calculated (Franceschini 2023). Establishing an MCID value may not be possible for FoF outcomes reported in this review, as FoF scale items may necessitate varying weightings for individuals and populations, considering that the significance of fear, concern, or worry about falling can differ across different situations (such as bathing, showering, or walking around the neighbourhood) (Kendrick 2014). Of note, Ellmers 2023, in one study on a special population, where people had cervical compressive myelopathy, estimated an MCID at 5.5 points using an anchor‐based calculation method and 10 using a distribution‐based method.

There was limited reporting of our secondary outcomes, particularly concerning adverse effects. Occurrence of falls was the most reported secondary outcome, in only eight of 11 studies for quantitative synthesis, whereas only one study reported on activity avoidance. Considering that FoF is defined as a "lasting concern about falling that leads to an individual avoiding activities that he/she remains capable of performing", this relevant measurement seems to be under‐reported.

Quality of the evidence

The GRADE approach assessed the certainty of evidence from moderate to very low across the different outcomes, as presented in Table 1. Several studies did not publish sufficient data to fully assess their risk of bias, resulting in uncertainty across many domains. Some authors responded to requests for additional information allowing for improving the certainty of the risk of bias (Freiberger 2012; Freiberger 2013; Gitlin 2006; Resnick 2008; Tennstedt 1998; Zijlstra 2009).

The body of evidence for all outcomes was downgraded due to serious risk of bias. Blinding of participants to their intervention groups was not possible, due to the nature of the psychological interventions and studies used self‐rated questionnaires for measuring outcomes, resulting in a high risk of performance and detection bias.

Serious inconsistency additionally lowered the certainty of evidence in most outcomes. Interventions varied significantly in duration, intensity, and design (e.g. AMB‐based, group versus individual, CBT only versus CBT with exercise). This was demonstrated by the presence of statistical heterogeneity between pooled studies. The low level of certainty for FoF sustainability of effects up to six months' postintervention was due to inconsistency between the CBT with and without exercise subgroups having substantial heterogeneity.

No outcomes were downgraded due to indirectness or publication bias. Most studies included in the review compared their intervention to a placebo, except for Wetherell 2018, who compared their intervention to a group receiving education. Although no studies reported adverse effects, we do not consider their omission a failure to report all results, but rather outcomes that are difficult to measure due to the nature of CBT interventions.

Except for postintervention effects of FoF interventions on activity avoidance, which showed low‐certainty evidence, all other secondary outcomes showed very low‐certainty evidence. This is likely due to the small number of studies and a low number of participants included in the analysis investigating secondary outcomes.

Potential biases in the review process

This review was designed to reduce the risk of bias during the review process. Despite a thorough search of literature, we could not rule out the possibility that we may not have included some eligible studies, such as unpublished studies or studies published in other languages not covered by the search. By including articles published only in English and German, there is potential for language bias. However, interventions for reducing FoF is not language specific, as is the case for traditional Chinese medicine. All relevant studies are expected to be covered in English and German publications. Therefore, we expect that language bias is small (if at all) and would not affect conclusions.

To reduce the risk of publication bias, we searched several databases for relevant studies for inclusion. We also searched reference lists of included studies for additional studies not identified in the original search strategy. However, this did not produce any new studies for inclusion.

To reduce the risk of reporting bias, when there were insufficient data available in the published manuscripts, we contacted study authors. This allowed outcome data on one occasion to be included in the study (Freiberger 2013).

Although the Cochrane guidelines and methods were strenuously followed during the development of the protocol, biases can still be present, affecting the procedures that are followed throughout the review. For example, we adhered to a strict definition of CBT that may have excluded studies using therapeutic approaches similar to our definition, but not entirely fulfilling our definition, such as some forms of motivational interviewing, or self‐help CBT (Lim 2023). Additionally, many studies during the review process had partial or no descriptors of speaking components of their intervention, which could not be verified due to the absence of study protocols, or intervention manuals. Therefore, it is possible that CBT interventions could have been missed due to a lack of adequate description in their titles, abstracts, and publications.

Deviations from the protocol may represent a source of potential bias. The review process limited the language of the review to studies published in English and German. This may exclude studies published in other languages, and as such not fully represent the entirety of currently available literature. Post‐hoc subgroup analyses were added. Post‐hoc analyses should be avoided due to the risk of "data mining" (Deeks 2022). This review, in trying to keep within the objectives of this review, selected the post‐hoc criteria to ensure that no important elements were missed. Additionally, we did not report sustainability of effects of secondary outcomes. Data for our secondary outcomes were limited by a smaller number of studies and participants. Because of the limited data, we judged that it was more meaningful to report immediate postintervention effects rather than explore sustainability of effects. We could not rule out the possibility that we introduced reporting biases owing to this decision. However, we consider the risk for bias minimal.

Wetherell 2018 used an active control group; FoF education. This study was retained in this review as education for FoF was deemed to be a sham treatment; education, although a part of CBT, does not on its own fulfil the criteria to be considered CBT or a therapeutic from a CBT perspective. This is supported by Rucker 2006, which found no reduction in FoF from FoF education as compared to usual care. However, although our interpretation of FoF education is that it is non‐therapeutic, as an active group, it may have the potential to influence FoF outcomes. For this reason, we conducted a post‐hoc sensitivity analysis by removing Wetherell 2018 from the analyses. Despite its removal, there was no change in the results.

Agreements and disagreements with other studies or reviews

For the primary outcome, this review concurs with similar reviews of CBT interventions for reducing FoF. One meta‐analysis conducted by Liu 2019 identified CBT‐only interventions to reduce FoF. It identified six trials involving 1626 participants. The meta‐analysis also included Huang 2016, investigating CBT interventions for reducing FoF in nursing home residents, and Liu 2014 who used an intervention known to reduce FoF as the control group (Tai Chi only versus Tai Chi with a CBT component). They did not include Reinsch 1992. Despite these variations, their results remained congruent with the results found in this review: FoF outcomes measured using the FES‐I are reduced over all time points (immediate postintervention, eight weeks, and two months) for both the Tai Chi‐only and Tai Chi with CBT interventions. In contrast to Liu 2019, our review reduced the risks of bias by including an additional study, and avoiding potential indirectness by including Liu 2014 in the pooled estimate.

A second meta‐analysis was conducted by Chua 2019 of CBT with and without exercise interventions for reducing FoF. Their meta‐analysis identified 15 studies for quantitative synthesis consisting of 3599 older community‐dwelling participants aged 60 years and above. They included additional studies that we excluded from this review, as we did not consider them to be CBT interventions but education. They also included Liu 2014 in their pooled estimate. In contrast, their meta‐analysis did not include Reinsch 1992, Resnick 2008, or Wetherell 2018. However, their results remained congruent with the results of this study: small effects in favour of CBT interventions immediately postintervention (SMD −0.28, 95% CI −0.35 to −0.21; 3165 participants, favoured intervention), and sustainability of effects up to six months (SMD –0.32, 95% CI –0.49 to –0.15; 1360 participants, favoured intervention) and more than six months (SMD –0.30, 95% CI –0.45 to –0.14; 1403 participants, favoured intervention). By using a stricter definition of CBT, our review provides a clearer understanding of the effects of CBT interventions, as well as providing subgroup analyses that Chua 2019 does not.

One related Cochrane Review exists for exercise‐only interventions for reducing FoF (Kendrick 2014). The population, comparison, and outcomes were similar to this review. The pooled effect size for exercise interventions was also small. They reported a small effect size for immediate postintervention effects (SMD 0.37, 95% CI 0.18 to 0.56, favoured intervention). Feng 2022 conducted a review of exercise‐only interventions "to evaluate the extent to which these interventions followed the exercise principles and reported exercise parameters, and quantify the effect of these interventions on reducing FOF." The review also found a small‐to‐moderate intervention effect in reducing FoF (SMD −0.34, 95% CI −0.44 to −0.23, favoured intervention). Caristia 2021 conducted a review assessing what type of exercise is associated with fall risk reduction in healthy adults aged 50 years and older. FoF, a secondary outcome, decreased with endurance exercises (Caristia 2021). We were unable to determine in our review if there were benefits to CBT interventions with or without exercise as opposed to exercise‐only interventions, and as such, there were insufficient data to assign individual patients to a specific intervention. The review investigated CBT interventions versus control, and not CBT interventions versus exercise interventions. Therefore, we could not extract comparisons between CBT and exercise interventions. Lenouvel 2021 suggested that the selection of a CBT intervention with or without exercise or exercise‐only intervention for reducing FoF could be guided by the patient frailty status; frail patients could benefit most from exercise as compared to non‐frail patients, and hence optimise their treatment.

Kruisbrink 2021 conducted a review of the relationship between characteristics and components of FoF interventions and intervention effectiveness, to determine which overarching characteristics of interventions and which components of interventions are effective in reducing FoF in community‐dwelling older people. These reviews demonstrated that FoF interventions in general were associated with a small‐to‐moderate reduction in FoF immediately postintervention. Body awareness, holistic exercises, meditation, and delivery of the intervention in the community setting were more effective at reducing FoF (Kruisbrink 2021). Given the aim of our review, we did not study the effects of characteristics and components of the CBT interventions with and without exercise. Additionally, many other factors can influence the effect of an intervention on FoF, for example, participant characteristics such as gender, comorbidity, and previous falls, and other factors related to the intervention, such as the quality of the intervention manuals, the quality of the delivery of the intervention components by the trainer, the quality of the adoption of the intervention components by the participants, and 'emotional and physical factors' enhancing the setting of the intervention and its atmosphere.

Although no study included in our review reported that there were adverse effects (harms) due to the intervention, Kempen 2011 conducted an observational study of the population from Zijlstra 2009 to examine the long‐term mortality effect of a multicomponent cognitive behavioural group intervention to reduce FoF and activity avoidance in community‐dwelling older people (Kempen 2011). They obtained mortality statuses of participants for several years following the end of intervention. There were no differences between groups allocated to intervention and control. However, when only intervention participants were included who participated in five or more sessions, there was a decrease in mortality after three years.

Of special note, two more‐recent studies investigated online self‐guided CBT interventions for FoF, that, due to lack of interactivity with a therapist, did not fulfil our inclusion criteria for CBT (Lim 2023; van Schooten 2021). Lim 2023 is an RCT that uses three modules (Managing Fear and Anxiety, Taking Charge of Worry, and Solving Problems) from the myCompass 2023 online self‐help CBT platform. These modules are considered relevant to FoF and are skill building. Participants have six weeks to complete the three modules. Follow‐up measures were at six weeks, six months, and 12 months. The trial showed high compliance with the modules, and low attrition rates, suggesting that online CBT is a feasible modality. However, it found no changes in FoF outcomes. van Schooten 2021 is an ongoing RCT. It adds a CBT component to the Standing Tall exercise intervention, using the myCompass 2023 platform given over 12 months.

Authors' conclusions

Implications for practice.

Cognitive behavioural therapy (CBT) with and without exercise interventions probably reduces fear of falling (FoF) in older people living in the community immediately after the intervention, may sustain improvements up to six months after intervention, and probably sustains improvements beyond six months. The standardised mean differences (SMDs) calculated for immediate postintervention effects, sustainability up to six months, and sustainability more than six months correspond to a reduction on the Falls Efficacy Scale – International (FES‐I) scale of 2.2 (95% confidence interval (CI) 3.4 to 1.0), 2.3 (95% CI 4.3 to 0.19), and 2.6 (95% CI 3.8 to 1.8) points, respectively, on a scale ranging from 16 (lowest level of FoF) to 64 (highest level of FoF). Subgroup analysis based on an intervention approach (CBT only versus CBT with exercise) shows similar SMDs, and as such, does not suggest that a particular approach is associated with greater improvement of FoF outcomes.

One of the main reasons why effect sizes are so small and changes potentially not clinically relevant, is that most studies in this review have not disentangled between maladaptive and non‐maladaptive processes. As pointed out in a comment on the recent World Fall Guidelines, concerns that reflect a realistic and appropriate appraisal of one's risk of falling could be well treated using structured exercise and balance training programmes (Montero‐Odasso 2022). However, if the concerns are indicative of maladaptive processes, it becomes essential to suggest psychological interventions such as CBT along with exercise.

Implications for research.

This review established moderate‐certainty evidence for CBT interventions for reducing FoF in older adults living in the community for immediate effects and sustainability of effects over six months' postintervention. However, the low certainty of evidence for sustainability up to six months creates a gap in the evidence. Further research may be justified to improve this. This can be accomplished through improved study design, reducing serious concern for imprecision (such as by ensuring sufficiently large study groups), and heterogeneity (such as reducing the risk for heterogeneity between individual studies, with the use of a standard measures, such as the FES‐I, and consistency in inclusion criteria, such as age). All studies had a risk of bias due to the nature of measuring FoF using self‐reported qualitative questionnaires. Recent advances in gait analysis, in part due to advances in sensor technology, have shown that FoF is associated with discrete gait characteristics that can be measured for changes over the course of treatment, therefore reducing the risk of detection bias (Ayoubi 2014; Lenouvel 2020).

A future direction of research may also consider not only relatively healthy older adults living in the community, but also those living in assisted living facilities and nursing homes, or special populations with different diseases or comorbidities (or both). Considering the significant proportion of the elderly population living in nursing homes or with comorbidities, this would allow for a better understanding of FoF interventions in a greater population. Relevant but still unclear aspects include gender‐related differences in FoF with higher prevalence of FoF in women despite comparable prevalence of falls (Pohl 2015), as well as gender‐specific aspects with respect to therapy initiation (Berger 2013) or treatment effects (Lim 2018). Future intervention providers should integrate information about gender‐specific factors that influence openness and acceptance of CBT interventions as well as FoF‐related aspects to target those who are likely to be most interested.

We adhere to a strict definition of CBT that is defined primarily through the element of interactivity, and as such requiring an in‐person presence. Recent trends in artificial intelligence may soon evolve to provide sufficient interactivity that the in‐person component is no longer required to fulfil this criterion of interactivity. Future studies could therefore take into consideration the rapid development of artificial intelligence programming in psychotherapy.

There were no differences found with certainty in the subgroup analyses. Future research should consider focusing on stratifying data according to baseline age or limiting their inclusion criteria to specific age groups of older adults. Additionally, future studies could investigate intervention characteristics, such as duration. Studies are needed that compare the dose‐effectivity of short‐ versus long‐term CBT interventions to better demonstrate the time to benefit effects.

This review cannot demonstrate preference for FoF between CBT interventions and exercise interventions. Future meta‐analyses could be designed to allow for this comparison. This would better improve our knowledge of the advantages of different approaches to treating FoF.

Reporting of this review's secondary outcomes amongst included studies was limited. This suggests an insufficiency of knowledge of secondary benefits to reducing FoF, such as fall rate, anxiety, depression, quality of life, and longevity.

Future studies should make careful consideration on terminology. The World Guidelines for Falls Prevention recommends using the terminology "concerns about falling" when making enquiries (Montero‐Odasso 2022). In addition, reporting on adverse effects should give careful attention to terminology, as there is a lack of standardisation and usage of adverse effects terminology (Deeks 2022). Future studies could report on adverse effects of CBT interventions, particularly considering that CBT has been shown to incur adverse effects (Parry 2016b; Strauss 2021). Furthermore, investigation of the sustainability of negative effects can help clarify how the adverse effects evolve following the end of treatment.

History

Protocol first published: Issue 3, 2021

Appendices

Appendix 1. Search strategies

The searches were run in three stages: the first search was run in March 2021 and top‐up searches were run in November 2021 and January 2023.

CENTRAL (CRS‐Web)

Search 1

#1 MESH DESCRIPTOR Accidental Falls AND CENTRAL:TARGET (1542)
#2 MESH DESCRIPTOR Fear AND CENTRAL:TARGET (1401)
#3 MESH DESCRIPTOR Phobic Disorders AND CENTRAL:TARGET (1190)
#4 ( fright* or fear* or afraid or phobia*):AB,EH,KW,KY,MC,MH,TI,TO AND CENTRAL:TARGET (11990)
#5 #2 OR #3 OR #4 (12347)
#6 #1 AND #5 (211)
#7 ((fear* or fright* or afraid or phobia*) NEAR5 fall*):AB,EH,KW,KY,MC,MH,TI,TO AND CENTRAL:TARGET (794)
#8 (ptophobia or post fall syndrome):AB,EH,KW,KY,MC,MH,TI,TO AND CENTRAL:TARGET (6)
#9 ("Falls Efficacy Scale" or "Mobility Efficacy Scale" or "Survey of Activities and Fear of Falling in the Elderly" or "University of Illinois at Chicago Fear of Falling Measure" or "SAFFE" or "UICFFM" or "Activities Specific Balance Confidence Scale" or "Confidence in Maintaining Balance Scale" or "CON‐Fbal"): AB,EH,KW,KY,MC,MH,TI,TO AND CENTRAL:TARGET (576)
#10 #6 OR #7 OR #8 OR #9 (1158)

Search 2 (top‐up search)

#11 16/03/2021_TO_11/11/2021:CRSCREATED AND CENTRAL:TARGET (76006)
#12 #10 AND #11 (92)

Search 3 (top‐up search)

#11 11/11/2021_TO_10/01/2023:CRSCREATED AND CENTRAL:TARGET (159146)
#12 #10 AND #11 (242)

MEDLINE (Ovid)

Search 1

1 Accidental Falls/ (25255)
2 Fear/ (32944)
3 *Phobic Disorders/ (7675)
4 (fright* or fear* or afraid or phobia*).tw. (97554)
5 or/2‐4 (109806)
6 1 and 5 (1405)
7 ((fear* or fright* or afraid or phobia*) adj5 fall*).tw. (2059)
8 ptophobia.tw. (2)
9 post fall syndrome.tw. (21)
10 ("Falls Efficacy Scale" or "Mobility Efficacy Scale" or "Survey of Activities and Fear of Falling in the Elderly" or "University of Illinois at Chicago Fear of Falling Measure" or "SAFFE" or "UICFFM" or "Activities Specific Balance Confidence Scale" or "Confidence in Maintaining Balance Scale" or "CON‐Fbal").tw. (935)
11 or/6‐10 (2787)
12 randomized controlled trial.pt. (525030)
13 controlled clinical trial.pt. (94095)
14 randomized.ab. (512874)
15 placebo.ab. (216118)
16 drug therapy.fs. (2289844)
17 randomly.ab. (353192)
18 trial.ab. (543638)
19 groups.ab. (2167207)
20 or/12‐19 (4941711)
21 exp animals/ not humans.sh. (4799766)
22 20 not 21 (4295798)
23 11 and 22 (974)

Search 2 (top‐up search)

24 (202103* or 202104* or 202105* or 202106* or 202107* or 202108* or 202109* or 202110* or 202111*).ed,dt. (1797486)
25 23 and 24 (168)

Search 3 (top‐up search)

24 (202111* or 202112* or 2022* or 2023*).ed,dt. (2360603)
25 23 and 24 (185)

Embase (Ovid)

Search 1

1 falling/ (42416)
2 fear/ (64015)
3 phobia/ (12015)
4 (fright* or fear* or afraid or phobia*).tw. (129486)
5 or/2‐4 (146330)
6 1 and 5 (1628)
7 ((fear* or fright* or afraid or phobia*) adj5 fall*).tw. (3042)
8 ptophobia.tw. (2)
9 post fall syndrome.tw. (32)
10 ("Falls Efficacy Scale" or "Mobility Efficacy Scale" or "Survey of Activities and Fear of Falling in the Elderly" or "University of Illinois at Chicago Fear of Falling Measure" or "SAFFE" or "UICFFM" or "Activities Specific Balance Confidence Scale" or "Confidence in Maintaining Balance Scale" or "CON‐Fbal").tw. (1408)
11 or/6‐10 (4131)
12 Randomized controlled trial/ (647655)
13 Controlled clinical study/ (466777)
14 Random*.ti,ab. (1634169)
15 randomization/ (90450)
16 intermethod comparison/ (269883)
17 placebo.ti,ab. (316157)
18 (compare or compared or comparison).ti. (513990)
19 ((evaluated or evaluate or evaluating or assessed or assess) and (compare or compared or comparing or comparison)).ab. (2266187)
20 (open adj label).ti,ab. (86178)
21 ((double or single or doubly or singly) adj (blind or blinded or blindly)).ti,ab. (235902)
22 double blind procedure/ (179713)
23 parallel group*1.ti,ab. (27169)
24 (crossover or cross over).ti,ab. (107644)
25 ((assign* or match or matched or allocation) adj5 (alternate or group*1 or intervention*1 or patient*1 or subject*1 or participant*1)).ti,ab. (348398)
26 (assigned or allocated).ti,ab. (410318)
27 (controlled adj7 (study or design or trial)).ti,ab. (370911)
28 (volunteer or volunteers).ti,ab. (249620)
29 trial.ti. (318749)
30 or/12‐29 (4905825)
31 (exp animal/ or animal.hw. or nonhuman/) not (exp human/ or human cell/ or (human or humans).ti.) (6175541)
32 30 not 31 (4249381)
33 11 and 32 (1437)

Search 2 (top‐up search)

34 2021*.dc,yr. (2140806)
35 33 and 34 (192)

Search 3 (top‐up search)

34 (2021* or 2022* or 2023*).dc,yr. (4377860)
35 33 and 34 (378)

CINAHL Plus

Search 1

S1 (MH “Accidental Falls”) (23,704)
S2 (MH "fear") (14,065)
S3 (MH "Phobic Disorders") (2,484)
S4 TI (fright* or fear* or afraid or phobia*) OR AB (fright* or fear* or afraid or phobia*) (41,269)
S5 S2 OR S3 OR S4 (47,142)
S6 S1 and S5 (1,353)
S7 7 TI ((fear* or fright* or afraid or phobia*) N5 fall*) OR AB ((fear* or fright* or afraid or phobia*) N5 fall*) (1,523)
S8 TI ptophobia or AB ptophobia (1)
S9 TI post fall syndrome OR AB post fall syndrome (6)
S10 TI ("Falls Efficacy Scale" or "Mobility Efficacy Scale" or "Survey of Activities and Fear of Falling in the Elderly" or "University of Illinois at Chicago Fear of Falling Measure" or "SAFFE" or "UICFFM" or "Activities Specific Balance Confidence Scale" or "Confidence in Maintaining Balance Scale" or "CON‐Fbal") OR AB ("Falls Efficacy Scale" or "Mobility Efficacy Scale" or "Survey of Activities and Fear of Falling in the Elderly" or "University of Illinois at Chicago Fear of Falling Measure" or "SAFFE" or "UICFFM" or "Activities Specific Balance Confidence Scale" or "Confidence in Maintaining Balance Scale" or "CON‐Fbal") (656)
S11 S6 OR S7 OR S8 OR S9 OR S10 (2,116)
S12 PT Clinical Trial (109,226)
S13 (MH “Clinical Trials+”) (315,777)
S14 TI clinical trial* OR AB clinical trial* (121,304)
S15 TI ( (single blind* or double blind*) ) OR AB ( (single blind* or double blind*) ) (50,549)
S16 TI random* OR AB random* (369,953)
S17 S12 OR S13 OR S14 OR S15 OR S16 (568,877)
S18 S11 AND S17 (463)

Search 2 (top‐up search)

S19 EM 20210316‐2021 (258,662)
S20 S18 AND S19 (25)

Search 3 (top‐up search)

S19 EM 20211111‐2023 (499,812)
S20 S18 AND S19 (56)

PsycINFO

Search 1

1 falls/ (3159)
2 fear/ (18562)
3 phobias/ (5365)
4 (fright* or fear* or afraid or phobia*).tw. (99810)
5 or/2‐4 (101104)
6 1 and 5 (488)
7 ((fear* or fright* or afraid or phobia*) adj5 fall*).tw. (872)
8 (ptophobia or post fall syndrome).tw. (7)
9 ("Falls Efficacy Scale" or "Mobility Efficacy Scale" or "Survey of Activities and Fear of Falling in the Elderly" or "University of Illinois at Chicago Fear of Falling Measure" or "SAFFE" or "UICFFM" or "Activities Specific Balance Confidence Scale" or "Confidence in Maintaining Balance Scale" or "CON‐Fbal").tw. (303)
10 or/6‐9 (1008)

Search 2 (top‐up search)

11 10 and 2021*.(sa_year) (37)

Search 3 (top‐up search)

11 10 and 2021*:2023*.(sa_year). (123)

AMED

Search 1

1 accidental falls/ (2445)
2 fear/ (558)
3 phobic disorders/ (180)
4 (fright* or fear* or afraid or phobia*).tw. (2349)
5 or/2‐4 (2418)
6 1 and 5 (249)
7 ((fear* or fright* or afraid or phobia*) adj5 fall*).tw. (299)
8 ("Falls Efficacy Scale" or "Mobility Efficacy Scale" or "Survey of Activities and Fear of Falling in the Elderly" or "University of Illinois at Chicago Fear of Falling Measure" or "SAFFE" or "UICFFM" or "Activities Specific Balance Confidence Scale" or "Confidence in Maintaining Balance Scale" or "CON‐Fbal").tw. (171)
9 or/6‐8 (429)

Search 2 (top‐up search)

10 2021*.up,yr. (8807)
11 9 and 10 (23)

Search 3 (top‐up search)

10 (2021* or 2022* or 2023*).up,yr. (19920)
11 9 and 10 (52)

WHO ICTRP

fall* AND fright* OR fall* AND fear* OR fall* AND afraid OR fall* AND phobia* (415)

ClinicalTrials.gov

(fall OR falls OR falling) AND (fright OR fear OR afraid OR phobia) | Interventional Studies (463)
(fall OR falls OR falling) AND (fright OR fear OR afraid OR phobia) | Interventional Studies First posted from 03/16/2021 to 11/11/2021 (41)
(fall OR falls OR falling) AND (fright OR fear OR afraid OR phobia) | Interventional Studies | First posted from 11/11/2021 to 01/10/2023 (99)

Appendix 2. Fear of falling measures

Measure	Studies that used the measure	Score interpretation	Description of measure
Falls Efficacy Scale (FES) (Tinetti 1990)	Gitlin 2006; Huang 2011	Lower scores represent lower concerns about falling (low FoF)	This self‐administered questionnaire is the original FoF scale for which there are many derivates. Items represent the 10 most important activities essential to independent living. This instrument has 10 items, rated on a 1 (not at all concerned) to 10 (very concerned) point Likert scale, with a total range between 10 and 100. Note: Gitlin 2006 added 3 items to this scale from the Activities‐Specific Balance Confidence Scale.
FES‐International (FES‐I) (Yardley 2005)	Dorresteijn 2016; Freiberger 2013; Parry 2016a; Wetherell 2018	Lower scores represent lower concerns about falling (low FoF)	A self‐administered questionnaire in its original version and an expanded, modified version of the FES. It uses 6 additional items representing various levels of challenging and social activities (Moore 2008). The FES‐I instrument has 16 items, rated on a 1 (not at all concerned) to 4 (very concerned) point Likert scale, with a total range between 4 and 64.
Modified FES (mFES) (Hill 1996)	Tennstedt 1998; Zijlstra 2009	Lower scores represent lower concerns about falling (low FoF)	This self‐administered questionnaire is an expanded version of the FES with an additional 4 items representing outdoor activities. This instrument has 14 items, rated on a 1 (not at all concerned) to 4 (very concerned) point Likert scale, with a total range between 4 and 64.
FES Swedish variant (FES‐S) (Hellström 1999; Hellström 2002)	Arkkukangas 2019	Lower scores represent lower concerns about falling (low FoF)	This self‐administered questionnaire is a translated version of the FES with 3 additional items, totalling 13 items rated on a 0 (not confident at all) to 10 (completely confident) point Likert scale, with a total range between 0 and 130.
Activities‐Specific Balance Confidence Scale (ABC) (Powell 1995):	Freiberger 2012	A high score represents a high level of physical functioning (low FoF)	This interviewer‐administered questionnaire has 16 items, rated on a 0% (no confidence) to 100% (completely confident) continuum, with a total score ranging between 0 and 100 (calculated by the total score divided by 16).
Single‐Item Instruments	Reinsch 1992; Resnick 2008	A high score represents high FoF	Reinsch 1992 used a single‐item question rated on a 5‐point scale regarding worry about falling has level of fear ranges from 1 = not at all worried to 5 = extremely worried. Resnick 2008 used a single‐item question asking respondents "How would you rate your fear of falling" on a scale ranging from 0 (no fear) to 4 (very afraid) (Resnick 2008).

FOF: fear of falling.

Appendix 3. Summary table, sensitivity analysis

Analysis	Outcomes (SMD)
Immediate postintervention effects
Removing the studies causing considerable heterogeneity (outliers)	−0.30 (95% CI −0.39 to −0.22; P = 0.65, I2 = 0%; 12 studies, 2228 participants)
Trials with high risks of bias (i.e. ≥ 3 domains of high risk of bias)	−0.17 (95% CI −0.32 to −0.01; P = 0.04, I2 = 50%; 9 studies, 1625 participants)
Comparing fixed‐effect and random‐effects estimates	−0.26 (95% CI −0.35 to −0.18; P = 0.03, I2 = 48%; 13 studies, 2357 participants)
Trials where FoF was not the primary outcome	−0.23 (95% CI −0.37 to −0.10; P = 0.01, I2 = 58%; 12 studies, 2266 participants)
Where non‐FES‐based questionnaires were used as measures	−0.31 (95% CI −0.40 to −0.20; P = 0.48, I2 = 0%; 9 studies, 2047 participants)
Clustered studies for potential ICC issues	−0.31 (95% CI −0.40 to −0.20; P = 0.48, I2 = 0%; 9 studies, 2047 participants)
Trials where active control groups were used	−0.25 (95% CI −0.40 to −0.10; P = 0.006, I2 = 63%; 9 studies, 2136 participants)
Sustainability of effects, up to 6 months' postintervention
Removing the studies causing considerable heterogeneity (outliers)	−0.31 (95% CI −0.44 to −0.17; P = 0.14, I2 = 36%; 8 studies, 1650 participants)
Trials with high risks of bias (i.e. ≥ 3 domains of high risk of bias)	−0.23 (95% CI −0.43 to −0.03; P = 0.003, I2 = 68%; 8 studies, 1470 participants)
Comparing fixed‐effect and random‐effects estimates	−0.27 (95% CI −0.36 to −0.17; P = 0.005, I2 = 63%; 9 studies, 1784 participants)
Trials where active control groups were used	−0.23 (95% CI −0.40 to −0.05; P = 0.004, I2 = 67%; 8 studies, 1744 participants)
Sustainability of effects, more than 6 months' postintervention
Removing the studies causing considerable heterogeneity (outliers)	−0.27 (95% CI −0.42 to −0.12; P = 0.20, I2 = 36%; 4 studies, 1117 participants)
Trials with high risks of bias (i.e. ≥ 3 domains of high risk of bias)	−0.24 (95% CI −0.39 to −0.08; P = 0.3, I2 = 19%; 4 studies, 871 participants)
Comparing fixed‐effect and random‐effects estimates	−0.28 (95% CI −0.40 to −0.17; P = 0.32, I2 = 14%; 5 studies, 1185 participants)

CI: confidence interval; FES: Falls Efficacy Scale; FoF: fear of falling; ICC: intracluster coefficient; SMD: standardised mean difference.

Appendix 4. Summary table, subgroup analysis, when outlier removed

Analysis	Subgroup	Outcomes (SMD)
Immediate postintervention effects
Age group	< 75 years	−0.11 (95% CI −0.54 to −0.32; 1 study, 90 participants)
	≥ 75 years	−0.31 (95% CI −0.40 to −0.22; P = 0.61, I2 = 0%; 9 studies, 1962 participants)
Type of control group	Placebo control	−0.32 (95% CI −0.41 to −0.23; P = 0.45, I2 = 0%; 8 studies, 2007 participants)
	Usual care	−0.13 (95% CI −0.42 to −0.16; P = 0.94, I2 = 0%; 4 studies, 221 participants)
Interventions based on AMB	AMB‐based studies	−0.31 (95% CI −0.45 to −0.17; P = 0.44, I2 = 0%; 2 studies, 783 participants)
	Non‐AMB studies	−0.30 (95% CI −0.41 to −0.19; P = 0.53, I2 = 0%; 10 studies, 1445 participants)
Group vs individual interventions	Group	−0.30 (95% CI −0.42 to −0.18; P = 0.65, I2 = 0%; 7 studies, 1125 participants)
	Individual	−0.30 (95% CI −0.43 to −0.18; P = 0.34, I2 = 11%; 5 studies, 1103 participants)
Sustainability of effects, up to 6 months' postintervention
Type of control group	Placebo control	−0.55 (95% CI −1.18 to −0.09; 1 study, 40 participants)
	Usual care	−0.30 (95% CI −0.43 to −0.16; P = 0.11, I2 = 42%; 7 studies, 1610 participants)
Interventions based on AMB	AMB‐based studies	−0.39 (95% CI −0.58 to −0.13; P = 0.68, I2 = 0%; 2 studies, 734 participants)
	Non‐AMB studies	−0.26 (95% CI −0.46 to −0.07; P = 0.11, I2 = 44%; 6 studies, 916 participants)
Group vs individual interventions	Group	−0.42 (95% CI −0.68 to −0.17; P = 0.19, I2 = 40%; 4 studies, 598 participants)
	Individual	−0.25 (95% CI −0.39 to −0.10; P = 0.27, I2 = 23%; 5 studies, 1052 participants)
Sustainability of effects, more than 6 months' postintervention
Interventions based on AMB	AMB‐based studies	−0.29 (95% CI −0.43 to −0.14; P = 0.45, I2 = 0%; 2 studies, 717 participants)
	Non‐AMB studies	−0.16 (95% CI −0.64 to 0.31; P = 0.04, I2 = 76%; 2 studies, 400 participants)
Group vs individual interventions	Group	−0.24 (95% CI −0.43 to −0.04; 1 study, 405 participants)
	Individual	−0.27 (95% CI −0.50 to −0.03; P = 0.11, I2 = 54%; 3 studies, 712 participants)

AMB: "A Matter of Balance"; CI: confidence interval; SMD: standardised mean difference.

Appendix 5. Mean difference outcomes, immediate postintervention effects for depression

Study	Intervention mean	Intervention SD	Intervention total	Control mean	Control SD	Control total	Mean difference IV, random, 95% CI
Parry 2016a	−1.16	3.54	151	0.1	2.65	163	−1.26 (95% CI −1.96 to −0.56)
Resnick 2008	0.41	0.79	56	0.79	1.1	34	−0.38 (95% CI −0.80 to 0.04)

CI: confidence interval; SD: standard deviation.

Appendix 6. Summary table, included study characteristics

Study	Total number	Mean Age	Number of arms	Design	Intervention types	Control type	Home treatment	Total duration and intensity of interventions	Therapist background	Baseline fall rates	FoF outcome measures
Arkkukangas 2019	124	> 75 years	3 (CBT only, CBT with exercise)	RCT	Individual, CBT with exercise	Usual care	Yes	6 hours, bimonthly treatment, for 12 weeks	Physiotherapists	43.2%	FES‐S
Dorresteijn 2016	389	> 70 years	2	RCT	Individual, CBT with exercise (AMB‐based)	Usual care	Yes	15.5 hours, bimonthly treatment, for 16 weeks	Nurses	62.2%	FES‐I
Freiberger 2012	280	> 70 years	4	Cluster‐RCT	Groups of up to 15 people, CBT with exercise (AMB‐based)	Usual care	No	32 hours, weekly treatment, for 16 weeks	Trained "fall prevention instructors"	26.1%	ABC
Freiberger 2013	378	> 65 years	2	RCT	Groups of 5–15 people, CBT with exercise	Usual care	No	16 hours, weekly treatment, for 16 weeks	Physiotherapists and sports scientists	52.3%	FES‐I
Gitlin 2006	319	> 70 years	2	RCT	Individual, CBT with exercise	Usual care	Yes	7.5 hours, monthly treatment, for 24 weeks	Physiotherapists and occupational therapists	—	FES
Huang 2011	186	> 60 years	3	RCT	Groups 8–12 people, CBT with exercise and CBT only	Usual care	No	8–12 hours, weekly treatment, for 8 weeks	Nurses	17.8%	FES
Parry 2016a	415	> 60 years	2	RCT	Individual, CBT only	Usual care	No	6.75 hours, weekly treatment, for 8 weeks	Licenced psychotherapists	—	FES‐I
Reinsch 1992	230	> 60 years	4 (CBT only, CBT with exercise, exercise only)	Cluster‐RCT	Groups 5–25 people, CBT with exercise and CBT only	Social contact and discussion group	No	1 hour, 3 days per week, for 1 year	—	27.1%	Single‐item instruments
Resnick 2008	166	> 60 years	2	Cluster‐RCT	Groups, CBT with exercise	Nutrition education	No	12–18 hours, biweekly treatment, for 12 weeks	Nutritionists or trained laypeople	—	Single‐item instruments
Tennstedt 1998	434	> 60 years	2	Cluster‐RCT	Groups, CBT with exercise	Social contact and discussion group	No	16 hours, biweekly treatment, for 8 weeks	Laypeople	24.9%	mFES
Wetherell 2018	42	> 65 years	2	RCT	Individual, CBT with exercise	Fall prevention education	Yes	8 hours, weekly treatment, for 8 weeks	Physiotherapists with supervision from psychotherapists	64.3%	FES‐I
Zijlstra 2009	540	> 70 years	2	RCT	Groups, CBT with exercise (AMB‐based)	Usual care	no	16 hours, weekly treatment, over 8 weeks	Nurses	55.6%	mFES

ABC: Activities‐Specific Balance Confidence Scale; AMB: "A Matter of Balance"; CBT: cognitive behavioural therapy; FES: Falls Efficacy Scale; FES‐I: FES‐International; FES‐IAB: FES International Avoidance Behaviour; FES‐S: Falls Efficacy Scale Swedish; FoF: fear of falling; mFES: Modified FES; RCT: randomised controlled trial; Single‐Item Instrument: two instruments.

Data and analyses

Comparison 1. Fear of falling: immediate postintervention.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1.1 Fear of falling: subgrouped according to intervention approach	11	2357	Std. Mean Difference (IV, Random, 95% CI)	‐0.23 [‐0.36, ‐0.11]
1.1.1 CBT only	3	472	Std. Mean Difference (IV, Random, 95% CI)	‐0.31 [‐0.56, ‐0.05]
1.1.2 CBT with exercise interventions	10	1885	Std. Mean Difference (IV, Random, 95% CI)	‐0.22 [‐0.36, ‐0.07]
1.2 Fear of falling: subgrouped according to mean age	10	2181	Std. Mean Difference (IV, Random, 95% CI)	‐0.23 [‐0.36, ‐0.09]
1.2.1 < 75 years	1	90	Std. Mean Difference (IV, Random, 95% CI)	‐0.11 [‐0.54, 0.32]
1.2.2 ≥ 75 years	9	2091	Std. Mean Difference (IV, Random, 95% CI)	‐0.23 [‐0.38, ‐0.09]
1.3 Fear of falling: subgrouped according to control	11	2357	Std. Mean Difference (IV, Random, 95% CI)	‐0.23 [‐0.36, ‐0.11]
1.3.1 Usual care	8	2136	Std. Mean Difference (IV, Random, 95% CI)	‐0.25 [‐0.40, ‐0.10]
1.3.2 Placebo	3	221	Std. Mean Difference (IV, Random, 95% CI)	‐0.13 [‐0.42, 0.16]
1.4 Fear of falling: subgrouped according to "A Matter of Balance" (AMB) versus non‐AMB based interventions	11	2357	Std. Mean Difference (IV, Random, 95% CI)	‐0.23 [‐0.36, ‐0.11]
1.4.1 Non AMB‐based interventions	8	1445	Std. Mean Difference (IV, Random, 95% CI)	‐0.30 [‐0.41, ‐0.19]
1.4.2 AMB‐based interventions	3	912	Std. Mean Difference (IV, Random, 95% CI)	‐0.11 [‐0.48, 0.25]
1.5 Fear of falling: subgrouped according to group versus individual interventions	11	2357	Std. Mean Difference (IV, Random, 95% CI)	‐0.23 [‐0.36, ‐0.11]
1.5.1 Individual intervention	5	1103	Std. Mean Difference (IV, Random, 95% CI)	‐0.30 [‐0.43, ‐0.18]
1.5.2 Group intervention	6	1254	Std. Mean Difference (IV, Random, 95% CI)	‐0.18 [‐0.39, 0.03]

Comparison 2. Fear of falling: sustainability of effects, up to 6 months' postintervention.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
2.1 Fear of falling: subgrouped according to intervention approach	8	1784	Std. Mean Difference (IV, Random, 95% CI)	‐0.24 [‐0.41, ‐0.07]
2.1.1 CBT only	2	404	Std. Mean Difference (IV, Random, 95% CI)	‐0.27 [‐0.47, ‐0.07]
2.1.2 CBT with exercise interventions	7	1380	Std. Mean Difference (IV, Random, 95% CI)	‐0.24 [‐0.46, ‐0.02]
2.2 Fear of falling: subgrouped according to control	8	1784	Std. Mean Difference (IV, Random, 95% CI)	‐0.24 [‐0.41, ‐0.07]
2.2.1 Usual care	7	1744	Std. Mean Difference (IV, Random, 95% CI)	‐0.23 [‐0.40, ‐0.05]
2.2.2 Placebo	1	40	Std. Mean Difference (IV, Random, 95% CI)	‐0.55 [‐1.18, 0.09]
2.3 Fear of falling: subgrouped according to AMB versus non‐AMB based interventions	8	1784	Std. Mean Difference (IV, Random, 95% CI)	‐0.24 [‐0.41, ‐0.07]
2.3.1 Non AMB‐based interventions	5	916	Std. Mean Difference (IV, Random, 95% CI)	‐0.26 [‐0.46, ‐0.07]
2.3.2 AMB‐based interventions	3	868	Std. Mean Difference (IV, Random, 95% CI)	‐0.19 [‐0.54, 0.17]
2.4 Fear of falling: subgrouped according to group versus individual interventions	8	1784	Std. Mean Difference (IV, Random, 95% CI)	‐0.24 [‐0.41, ‐0.07]
2.4.1 Individual intervention	5	1052	Std. Mean Difference (IV, Random, 95% CI)	‐0.25 [‐0.39, ‐0.10]
2.4.2 Group intervention	3	732	Std. Mean Difference (IV, Random, 95% CI)	‐0.25 [‐0.64, 0.15]

Comparison 3. Fear of falling: sustainability of effects, more than 6 months' postintervention.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
3.1 Fear of falling: subgrouped according to intervention approach	5	1185	Std. Mean Difference (IV, Random, 95% CI)	‐0.28 [‐0.40, ‐0.15]
3.1.1 Sustainability of effects, CBT only	1	314	Std. Mean Difference (IV, Random, 95% CI)	‐0.38 [‐0.60, ‐0.15]
3.1.2 Sustainability of effects, CBT with exercise	4	871	Std. Mean Difference (IV, Random, 95% CI)	‐0.24 [‐0.39, ‐0.08]
3.2 Fear of falling: subgrouped according to "A Matter of Balance" (AMB) versus non‐AMB based interventions	5	1185	Std. Mean Difference (IV, Random, 95% CI)	‐0.28 [‐0.40, ‐0.15]
3.2.1 Non AMB‐based interventions	2	400	Std. Mean Difference (IV, Random, 95% CI)	‐0.16 [‐0.64, 0.31]
3.2.2 AMB‐based interventions	3	785	Std. Mean Difference (IV, Random, 95% CI)	‐0.29 [‐0.43, ‐0.14]
3.3 Fear of falling: subgrouped according to group versus individual interventions	5	1185	Std. Mean Difference (IV, Random, 95% CI)	‐0.28 [‐0.40, ‐0.15]
3.3.1 Individual intervention	3	712	Std. Mean Difference (IV, Random, 95% CI)	‐0.27 [‐0.50, ‐0.03]
3.3.2 Group intervention	2	473	Std. Mean Difference (IV, Random, 95% CI)	‐0.24 [‐0.42, ‐0.06]

Comparison 4. Secondary outcomes.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
4.1 Activity avoidance: immediate postintervention	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
4.2 Occurrence of falls: immediate postintervention	5	1119	Risk Ratio (M‐H, Random, 95% CI)	0.96 [0.66, 1.39]
4.2.1 CBT only	1	90	Risk Ratio (M‐H, Random, 95% CI)	0.50 [0.21, 1.20]
4.2.2 CBT with exercise intervention	5	1029	Risk Ratio (M‐H, Random, 95% CI)	1.04 [0.70, 1.53]
4.3 Depression: immediate postintervention	2	404	Std. Mean Difference (IV, Random, 95% CI)	‐0.41 [‐0.60, ‐0.21]
4.3.1 CBT only	1	314	Std. Mean Difference (IV, Random, 95% CI)	‐0.40 [‐0.63, ‐0.18]
4.3.2 CBT with exercise	1	90	Std. Mean Difference (IV, Random, 95% CI)	‐0.41 [‐0.84, 0.02]
4.4 Anxiety: immediate postintervention	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected
4.5 Quality of life: immediate postintervention	4	701	Std. Mean Difference (IV, Random, 95% CI)	‐0.24 [‐0.51, 0.04]
4.5.1 CBT only	3	472	Std. Mean Difference (IV, Random, 95% CI)	‐0.04 [‐0.23, 0.14]
4.5.2 CBT with exercise	3	229	Std. Mean Difference (IV, Random, 95% CI)	‐0.47 [‐0.98, 0.05]

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Arkkukangas 2019.

Study characteristics
Methods	3‐arm RCT Randomisation performed by an independent statistician who transferred the sequences to consecutively numbered envelopes.
Participants	Country: Sweden n = 175, at‐home sessions Female: 70% Mean age (years): 83 (SD 4.7) Inclusion criteria: aged ≥ 75 years, able to walk independently at home, able to understand written and oral information in the Swedish language Exclusion criteria: MMSE < 25; ongoing, regular physical therapy treatment due to injury or illness (or both) or being in terminal care
Interventions	CBT with exercise Arm 1: OEP (n = 61): a series of 17 strength and balance exercises Arm 2: OEP with MI (n = 58): MI consisted of open‐ended questions, affirmations, reflective listening, and summaries (Miller 1991). The instruction session was calculated to last approximately 1 hour and consisted of the OEP combined with MI. The intervention groups had 6 home visits, with the first 5 during the first 3 months. 3 times weekly exercises at home (1 hour), with walks between sessions (30 minutes). All sessions lasted about 1 hour. 3 telephone calls. Arm 3: control (n = 56): standard care. No treatment, only surveillance.
Outcomes	FES‐S; SPPB; handgrip strength; EQ‐5D; EQ‐5D VAS; Frädlin‐Grimby Activity Scale; number of falls
Notes	Study start date: November 2012, end date: December 2013. No conflicts of interest declared. Source of funding: the National Swedish Board of Health and Welfare, Grants for the County of Västmanland. Regional Research Fund for Uppsala and Örebro region, Sweden. Research and Development Department in the Community of Eskilstuna, Sweden. Protocol published in 2014. Intermediary results published in 2017.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Used computer‐generated blocked allocation.
Allocation concealment (selection bias)	Low risk	Randomisation performed in blocks to minimise the risk of revealing treatment group allocation.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Blinding of participants not possible due to nature of intervention.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Assessors blinded, but outcomes were self‐reported by participants who knew their group allocation.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Few participants lost to follow‐up.
Selective reporting (reporting bias)	Low risk	Protocol published 2014. Study was congruent with protocol.
Other bias	Low risk	Appeared free of other sources of bias.

Dorresteijn 2016.

Study characteristics
Methods	RCT Randomisation by external agency blinded to participant characteristics
Participants	Country: the Netherlands n = 389 (intervention: 194, control: 195), at‐home sessions Female: 72.3% Mean age (years): control: 78.25 (SD 5.3), intervention: 78.35 (SD 5.4) Inclusion criteria: at least sometimes concerned about falls and associated activity avoidance, and valued their perceived general health as fair or poor assessed with 1 item of the Medical Outcomes Study‐20 (MOS‐20), and signed the informed consent Exclusion criteria: confined to bed; wheelchair dependent; waiting for nursing home admission; experienced substantial hearing, vision, or cognitive impairments.
Interventions	CBT with exercise Arm 1: home‐based format of AMB; 7 individual sessions, including 3 home visits and 4 telephone contacts Each session had a predefined theme; concerns about falls; thoughts about falling; physical exercise; asserting oneself; overcoming personal barriers; safe behaviour; managing concerns about falls. Each session was similarly structured; review of the previous session (except the first session), a discussion of the main theme, the formulation of a personalised action plan related to the discussed theme. In session 5, participants were guided to safely execute a daily activity they were afraid to perform independently (exposure in vivo) Delivered by community nurses using detailed manuals. Arm 2: usual care
Outcomes	FES‐I; FES‐IAB; GARS; number of falls (fall diary)
Notes	Study start date: March 2009, end date: March 2011 (end of intervention: December 2009, with 15 months of observation). No conflicts of interest declared. Source of funding: ZonMw, the Netherlands Organization for Health Research and Development (grant 120610001) and for author Vlaeyen J, a grant from the Research Foundation, Flanders, Belgium (FWO Vlaanderen).
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation conducted by an external agency blinded to participant characteristics.
Allocation concealment (selection bias)	Low risk	Outcome assessors blinded to the allocation.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Blinding of participants not possible due to nature of intervention.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Assessors blinded, but outcomes were self‐reported by participants who knew their group allocation.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Study authors conducted intention‐to‐treat analysis and imputed missing data.
Selective reporting (reporting bias)	Low risk	Trial registry protocol available.
Other bias	Low risk	Appeared free of other sources of bias.

Freiberger 2012.

Study characteristics
Methods	4‐arm RCT Randomisation by a third party not involved in the study using a computerised random‐number generator.
Participants	Country: Germany n = 122 Female: 122 (43.6%) Mean age (years): 76.1 (SD 4.1) Inclusion criteria: aged ≥ 70 years, having fallen in the past 6 months or FoF, provided signed informed consent, and completed baseline assessment Exclusion criteria: unable to ambulate independently, cognitive impairment (DSST score < 25)
Interventions	CBT with exercise Group‐based (< 15 participants), 2 × 1‐hour sessions per week, over 16 weeks, at an institution (Institute of Sports Sciences, Erlangen, Germany) Delivered by "fall prevention instructors" (trained by physiotherapist or university‐trained sport scientist) Arm 1: Strength and Balance group (n = 63); progressive exercises for upper and lower body strength (standing weight‐bearing exercises with dumbbells and challenging balance exercises), motor co‐ordination training, ball games, obstacle courses, sensory training (standing and walking with eyes closed or on unstable surfaces), training in the perception of centre of gravity. Individual progression in strength training was monitored and adjusted according to the Self‐Perceived Exertion Scale. Arm 2: Fitness group (n = 64); endurance training (walking with change of pace and direction, Nordic walking) in addition to Strength and Balance group activities. Individual progression was monitored and adjusted according to the Self‐Perceived Exertion Scale. Arm 3: Multifaceted group (n = 73); AMB component (fall risk education; elements addressed were in particular attitudes about falls, thoughts and concerns about falling and their effects regarding feelings and behaviour, and recognising potential environmental fall hazards) and cognitive training (exercises on concentration, information processing speed, and short‐term memory) in addition to the Strength and Balance group activities. Arm 4: control group (n = 80)
Outcomes	ABC; 2 subscales of the perceived Consequences of Falling scale (Loss of Functional Independence, Damage to Identity); DSST; Single‐item FoF (yes/no); falls; number of days walking per week; TUG; modified Romberg Test; chair rise test; self‐selected normal and fast walking speed over a 10‐m distance
Notes	Study start date: 2003, end date: 2006. No conflicts of interest declared. Source of funding: The Robert Bosch Foundation and Siemens Health Insurance financially supported this project. Protocol published.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated randomisation.
Allocation concealment (selection bias)	Low risk	Randomisation performed in blocks to minimise the risk of revealing treatment group allocation.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Blinding of participants not possible due to nature of intervention.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Assessors blinded, but outcomes were self‐reported by participants who knew their group allocation.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Missing data balanced between groups by modified intention‐to‐treat analysis using available data.
Selective reporting (reporting bias)	Low risk	Study protocol published.
Other bias	Low risk	Appeared free of other sources of bias.

Freiberger 2013.

Study characteristics
Methods	Cluster‐RCT Blockwise randomisation list for both co‐ordination centres, by a statistician who was otherwise uninvolved in the study at the time.
Participants	Country: Germany n = 325 (intervention: 184, control: 141) Female: intervention: 77.5%, control: 72.4% at baseline Inclusion criteria: aged ≥ 65 years, reporting ≥ 1 falls in the past year or FoF or physical fall risk obtained via specific fall risk assessment (or a combination of these), being mobile (able to stand alone and walk alone or with assistive device) Exclusion criteria: wheelchair dependent
Interventions	CBT with exercise Arm 1: group intervention with 5–15 participants, 16 supervised sessions, once weekly for 60 minutes. Group sessions focused on CBT twice in month 2, twice in month 3, and once in month 4, total 5 times. Starting from week 5, participants added ≥ 1 unsupervised session for total number of session of 28. Supervised session structure; 5‐minute discussion to introduce the topic of the session and address participants' well‐being and questions, followed by a 10‐minute warm‐up phase, leading to a 30‐ to 40‐minute conditioning period, followed by a 5‐ to 10‐minute cooling down and closing phase with relaxation and discussion between the participants and instructors about the experience. Group sessions focused on CBT twice in month 2, twice in month 3, and once in month 4, total 5 times Exercise programme used only bodyweight, no additional equipment Arm 2: control group: standard care Delivered by physiotherapist or university‐trained sport scientist
Outcomes	FES‐I; CST; TUG; modified Romberg test
Notes	Study start date: 2011 (not specified explicitly), end date: 2012 (study duration 12 months). No conflicts of interest were declared. Source of funding: The Bavarian State Ministry of the Environment and Public Health (Gesund. Leben. Bayern.) (Grant number: LP 00110, Pr Nr 09‐10). Additional study data not available in study manuscript were obtained following contact with author (Ellen Freiberger) Protocol published. Intervention developed from Freiberger 2012.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Blockwise randomisation list for both co‐ordination centres, by a statistician who was otherwise uninvolved in the study at the time.
Allocation concealment (selection bias)	High risk	No blinding of staff for participant allocation and group allocation was known before participants were included.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Blinding of participants not possible due to nature of intervention.
Blinding of outcome assessment (detection bias) All outcomes	High risk	No blinding of assessors and outcomes were self‐reported by participants who knew their group allocation.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Outcome data based on > 80% of intervention group.
Selective reporting (reporting bias)	Low risk	Study protocol published in trial registry
Other bias	Low risk	Additional risk of bias domain for cluster RCTs Recruitment: no concerns identified Baseline imbalance: no concerns identified Loss of clusters: no concerns identified Incorrect analysis: clustering was accounted for in the study's analysis Comparability with individually randomised trials: no concerns identified

Gitlin 2006.

Study characteristics
Methods	RCT Randomisation: stratification (white people, non‐white people, and living alone or with others), randomised within each of 4 strata using random permuted blocks. The blocking number remained unknown. Randomisation lists and 4 sets of randomisation were prepared using double, opaque envelopes.
Participants	Country: USA n = 319 (intervention: 160, control: 159) Female: intervention: 82.5%, control: 81.1% Mean age (years): intervention: 79.5 (SD 6.1), control: 78.5 (SD 5.7) Inclusion criteria: aged ≥ 70 years, cognitively intact (MMSE > 23), English speaking Exclusion criteria: receiving home care, reported the need for help or difficulties with 2 iADLs or ≥ 1 ADLs, non‐community living
Interventions	Arm 1: individual, at home; 5 occupational therapy contacts (4 × 90‐minute visits and 1 × 20‐ minute telephone contact), 1 physical therapy visit, then 3 occupational therapist calls (reinforce the use of intervention derived strategies and generalise these strategies to new problem areas) over following 6 months. Semi‐structured clinical interview to identify and prioritise problem areas. Problem‐solving to identify behavioural and environmental contributors to performance difficulties. Control‐enhancing strategies through cognitive (problem‐solving, reframing), behavioural (pace self, sit instead of stand to perform tasks), and environmental modifications. Delivered by occupational therapists Arm 2: usual care
Outcomes	ADLs; iADLs; mobility/transfer; FES‐I + 3‐items of ABC (confident walking up/down stairs, bending/picking up slipper from floor, and getting into/out of car without falling).
Notes	Study start date: 2000, end date: 2003. No conflicts of interest declared. Source of funding: USA National Institute on Aging Grant R01 AG13687.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation conducted by a statistician and generated by a computer, in random permuted blocks that remained unknown to others.
Allocation concealment (selection bias)	Low risk	Randomisation performed in blocks and stored in double opaque envelopes.
Blinding of participants and personnel (performance bias) All outcomes	High risk	No blinding of participants.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Assessors blinded, but outcomes were self‐reported by participants who knew their group allocation.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Outcome data based on > 80% of intervention group.
Selective reporting (reporting bias)	Unclear risk	No protocol for the study available.
Other bias	Low risk	Appeared free of other sources of bias.

Huang 2011.

Study characteristics
Methods	3‐arm RCT Randomisation used a computer‐developed random table to randomly assign participants to 3 intervention groups
Participants	n = 186, CBT with exercise intervention arm; 62, CBT only arm; 62, control; 62 Female: CBT with exercise: 64.5%, CBT only: 54.8%, control: 56.4% Inclusion criteria: aged ≥ 60 years, mentally intact (SPMSQ score > 6 for illiterate individuals, > 7 for those with 6 years of education, > 8 for those with > 6 years of education), resident in the community, and able to communicate in Mandarin or Taiwanese Exclusion criteria: had an artificial leg or leg brace, unstable health problems, terminally ill.
Interventions	CBT (FoF management model) with exercise (Tai Chi) Arms 1 and 2: groups of 8–12 participants, weekly sessions for 8 weeks, 60–90 minutes. CBT component's main strategy was restructuring misconceptions to promote a view of fall risk and FoF as controllable. Each session covered the following topics: 1. introduction; 2. associations with falls or FoF; 3. participant's point of view of FoF (positive and negative aspects about the topic); 4. strategies to manage FoF and family support; 5. implementation in the participants' daily life; and 6. problem‐solving (during a fall learning how to fall, stand up, and call for help). Tai Chi consisted of 10 positions derived from the Yang style. Participants in groups of 10–16 had lessons 5 times a week, for 8 weeks. Each session of Tai Chi began with warm‐up exercises (10 minutes), followed by teaching and practicing individual forms of the Tai Chi programme (45 minutes), and ended with cool‐down exercises (5 minutes). Participants in the CBT with Tai Chi intervention needed to complete the Tai Chi at least 3 times a week for 8 weeks, and the CBT intervention all 8 sessions. Given by geriatric and community health nurses (CBT component) and professional Tai Chi instructors (exercise component) Arm 3: usual care
Outcomes	FES‐I; GFFM; number of falls; Tinetti Mobility Scale; ISSB; WHOQOL‐BREF (quality of life)
Notes	Study start date: March 2007, end date: December 2007. No conflicts of interest declared. Sources of financing: National Science Council, Taiwan (Grant number: NSC97‐2314‐B‐182‐031‐MY3).
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Participants randomly assigned to a group with a computer‐developed random table.
Allocation concealment (selection bias)	Low risk	Allocation concealed.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Blinding of participants not possible due to nature of intervention.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Assessors blinded, but outcomes were self‐reported by participants who knew their group allocation.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Outcome data based on > 80% of intervention group.
Selective reporting (reporting bias)	Unclear risk	Study protocol not published.
Other bias	Low risk	Appeared free of other sources of bias.

Parry 2016a.

Study characteristics
Methods	RCT Randomisation: a computer‐generated blocked allocation, and stratified by site, patient gender, baseline score on the numeric rating scale for pain when walking (0 vs 1–10) and whether or not the patient had been referred for strength and balance training
Participants	Country: UK n = 415, intervention: 210, control: 205 Female: intervention: 75.8% (SD 8.5%), control: 75.3% (SD 8.6%) Mean age (years): intervention: 69.5, control: 70.7 Inclusion criteria: FES‐I score > 23, aged ≥ 60 years, community‐dwelling Exclusion criteria: cognitive impairment (MMSE < 24/30), life expectancy < 1 year or unlikely for any other reason to be unable to complete 1‐year follow‐up, requiring psychosocial interventions that are unrelated to FoF, current involvement in other investigational studies or trials, or involvement within 30 days prior to study entry, had taken part in previous phase of the study.
Interventions	CBT‐only intervention, individual, weekly (45 minutes with 15‐minute preparation time) for 8 weeks, with a 6‐month booster session. Arm 1: standard CBT interventions for identified problems (the 3 P's (predisposing, precipitating, and perpetuating) model) were described and practised: graded exposure for anxiety; activity monitoring, graded activity and behavioural activation for pain, fatigue, and low mood; and sleep management for fatigue. For other recurrent issues, healthcare assistant were taught specific cognitive techniques, such as identifying thoughts; making explicit the links between thoughts, feelings, and behaviour; thought diaries; thought challenging; and cost–benefit analysis for catastrophic and otherwise unhelpful cognitions. Delivered by healthcare assistants Arm 2: standard care for falls service
Outcomes	FES‐I; Falls (falls diary); HADS; Numeric Rating Scale for pain when walking; Numeric Rating Scale for FoF when walking; MMSE; WHOQOL‐OLD; EQ‐5D (5L); SF‐36; 11‐item De Jong Gierveld Loneliness Scale; LSNS‐6; Social Participation Questionnaire; SPPB; functional reach; handgrip strength; adverse events
Notes	Study start date: 3 July 2012, end date: 31 January 2015. No conflicts of interest declared. Sources of funding: NIHR Health Technology Assessment programme
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐developed random table randomised participants.
Allocation concealment (selection bias)	Low risk	Group allocation concealed from a research assistant who was only responsible for participant recruitment and group allocation in this study.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Blinding of participants not possible due to nature of intervention.
Blinding of outcome assessment (detection bias) All outcomes	High risk	The research team, including those collecting outcome data and the statistician analysing data, were not blinded, and outcomes were self‐reported by participants who knew their group allocation.
Incomplete outcome data (attrition bias) All outcomes	High risk	Missing data were imputed with intention‐to‐treat analysis.
Selective reporting (reporting bias)	Low risk	Protocol published in manuscript.
Other bias	Low risk	Appeared free of other sources of bias.

Reinsch 1992.

Study characteristics
Methods	4‐arm cluster‐RCT 16 senior centres were randomly assigned to 1 of 4 treatment groups (exercise, CBT, CBT with exercise, discussion control), and all participants at a centre were given the same intervention.
Participants	Country: USA n = 230 Females: number not reported Inclusion criteria: community‐dwelling people aged ≥ 60 years Exclusion criteria: did not meet inclusion criteria
Interventions	Arm 1: CBT; 1 hour, 1 day per week, for 1 year. Each class meeting covered a health and safety curriculum to prevent falls, relaxation training to lower tension and fear, and videogame playing to improve reaction time. There was a CBT protocol. Arm 2: CBT with exercise; 1 hour, 3 days per week, for 1 year; 1 class meeting per week followed the CBT protocol and 2 meetings focused on exercise but also included relaxation training and discussion of safety topics as they arose. Arm 3: exercise; Stand‐up/Step‐up program: requiring a specified number of "stand‐ups" from a sitting position and "step‐ups" onto a 6‐inch‐high stepping stool. Warm‐up and cool‐down consisted of stretching and movement to music. Arm 4: control; 1 hour, 1 day per week, for 1 year; health and discussion topics not specifically related to fall prevention but of interest to older people were the focus of the discussion control group meetings.
Outcomes	Falls; injury; strength and balance; FoF (single‐item measure 1–5); self‐rated health
Notes	Start date: not specified, end date: 12 months after start. No conflicts of interest declared. Source of funding: NIH Grant #AG07350, by the AARP Andrus Foundation, and by the Roosevelt Warm Springs Foundation.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation was to centres, rather than individuals, as there was only space for 1 class at each centre. How randomisation was done was not stated.
Allocation concealment (selection bias)	Unclear risk	Although centres were randomly assigned treatments, it is not clear if inclusion of participants was conducted prior to randomisation procedure.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Blinding of participants not possible due to nature of intervention.
Blinding of outcome assessment (detection bias) All outcomes	High risk	Unclear if assessors were blinded and outcomes were self‐reported by participants knowing their group allocation.
Incomplete outcome data (attrition bias) All outcomes	High risk	Outcome data for FoF (secondary outcome) based on < 80% of intervention group.
Selective reporting (reporting bias)	Unclear risk	Protocol not published.
Other bias	Unclear risk	Additional risk of bias domain for cluster RCTs Recruitment: control group had disproportionate ratio of females to males. Baseline imbalance: no concerns identified. Loss of clusters: no concerns identified. Incorrect analysis: study did not adjust for clustering in their analysis. Comparability with individually randomised trials: no concerns identified.

Resnick 2008.

Study characteristics
Methods	Cluster‐RCT Sites were randomised by geographic cluster. A coin was tossed to determine allocation
Participants	Country: USA n = 166 Female: intervention: 79%, control: 83% Inclusion criteria: aged ≥ 60 years; blood pressure < 200 mmHg systolic and < 100 mmHg diastolic; heart rate 60–120 beats per minute; and no known recent (within the past 6 months) history of myocardial infarction, stroke, or new irregular heartbeat Exclusion criteria: if they did not obtain a signed note from their primary health care provider acknowledging their participation.
Interventions	Arm 1: CBT with exercise (Senior Exercise Self‐efficacy Project (SESEP)), 3 sessions/week, 30‐minute efficacy enhancing component, twice‐weekly nutrition education, daily home exercises (30 minutes) for 12 weeks An "efficacy enhancing component" consisted of enactive 1. enactive mastery experiences, involving successful performance of the activity of interest; 2. verbal persuasion or verbal encouragement by a credible source, consisting of messages that one is capable of performing the activity; 3. vicarious experience, which involves seeing individuals similar to oneself perform the activity; 4. reinterpretation/control of physiological and affective states physiological and affective states such as joy, pain, fatigue, or anxiety associated with or experienced during the activity; and 5. promotion of positive outcome expectancies, by education about health benefits of exercise. Participants were given a booklet containing stretching, resistance, and aerobic exercise activities described in a recommended exercise programme, with images and brief descriptions of how to perform each exercise. Arm 2: control; nutrition course of equal intensity Delivered by nutritionists or trained laypeople
Outcomes	FoF evaluated by asking the participant to rate his or her FoF on a scale of 0–4; SEE; OEE; YPAS; SF‐12; GDS; Tinetti Scale; chair rise time; pain (0–10 scale)
Notes	Start and end dates not specified. No conflicts of interest declared. Sources of funding: NYC DOHMH, the New York State Department of Health, and the Robert Wood Johnson Foundation.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Coin toss used to randomise centres.
Allocation concealment (selection bias)	Low risk	Coin toss used to determine allocation between the Brooklyn sites and the South Bronx/Upper Manhattan sites.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Blinding of participants not possible due to nature of intervention.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Assessors blinded, but outcomes were self‐reported by participants who knew their group allocation.
Incomplete outcome data (attrition bias) All outcomes	High risk	Outcome data based on < 80% of intervention group.
Selective reporting (reporting bias)	Unclear risk	Protocol not available.
Other bias	Unclear risk	Additional risk of bias domain for cluster RCTs Recruitment: no concerns identified Baseline imbalance: no concerns identified Loss of clusters: no concerns identified Incorrect analysis: study did not adjust for clustering in their analysis Comparability with individually randomised trials: no concerns identified

Tennstedt 1998.

Study characteristics
Methods	RCT Unit of randomisation was the housing site. 40 sites recruited for participation and pair‐matched on the basis of number of units and percentage of ethnic minority residents, with 1 site in each pair randomly assigned to the intervention group and the other site to a placebo attention control group.
Participants	Country: USA n = 434 Female: 89.6% Age (years): 77.8 (SD 7.71) Inclusion criteria: aged ≥ 60 years; absence of any major physical or health condition; English speaking; and self‐reported restriction in activity due to fear of falling Exclusion criteria: inclusion criteria not fulfilled
Interventions	Arm 1: group CBT with exercise intervention (AMB), twice‐weekly, 2 hours sessions for 4 weeks The early sessions focused on changing attitudes and self‐efficacy prior to attempting changes in actual behaviour. A cognitive restructuring approach instilled adaptive beliefs. Varying activities promoted an adaptive conception of fear of falling with training exercises on how to shift from maladaptive (self‐defeating) to adaptive (motivating) cognitions. Behavioural contracts and goal‐setting regarding desirable changes (e.g. correcting identified home hazards; engaging in physical exercise; resuming a formerly restricted activity) were used. Strength training exercises (using wide elastic bands for resistance) were included in 6/8 sessions to instruct and encourage participants to continue them independently. Arm 2: control group had a single 2‐hour group session consisting of a didactic presentation regarding incidence and risk factors for falls, a video on home hazards that increase fall risk, and steps that can be taken to reduce risk, and a group discussion.
Outcomes	FES with 2 additional items; carrying bundles from the store and exercising; falls; SIP
Notes	Start date: October 1994, end date: July 1996 No conflicts of interest declared. Sources of funding: National Institute on Aging (Grant No. AG11669). Contacted authors but study data not retrievable.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Did not explain how randomisation was conducted.
Allocation concealment (selection bias)	Unclear risk	Due to cluster randomisation, group allocation appeared to be known prior to recruitment of participants.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Blinding of participants not possible due to nature of intervention.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Assessors blinded, but outcomes were self‐reported by participants who knew their group allocation.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Outcome data based on > 80% of intervention group.
Selective reporting (reporting bias)	Unclear risk	Protocol not published.
Other bias	High risk	Abstract misleading as only compliant group is discussed. Additional risk of bias domain for cluster RCTs Recruitment: no concerns identified Baseline imbalance: no concerns identified Loss of clusters: no concerns identified Incorrect analysis: study did not adjust for clustering in their analysis. The study reported only partial data. Comparability with individually randomised trials: no concerns identified

Wetherell 2018.

Study characteristics
Methods	RCT Randomisation was determined using a computer‐generated sequence created and held by a colleague with no other connection to the study; project co‐ordinator obtained the assignment after each participant completed the baseline assessment.
Participants	Country: USA n = 42 Female: intervention: 66.7%, control: 81.0% Age (years): intervention: 77.3 (SD 7.0), control: 78.5 (SD 7.8) Inclusion criteria: community‐dwelling people aged 65–91 years, FES‐I > 27 Exclusion criteria: high objective risk of falls or risk of injury: > 2 falls in the past year, requiring the assistance of another person to walk or transfer, orthostasis, history of osteoporotic fracture, body mass index ≤ 17, cognitive impairment as measured by 10 or more errors on the Blessed Orientation Memory and Concentration Test, or legally blind; history of schizophrenia or bipolar disorder, alcohol or drug abuse in the past 6 months, active suicidal ideation, or currently receiving physical therapy or psychotherapy.
Interventions	Arm 1: ABLE: individual at‐home CBT with exercise intervention; 8 weekly in home sessions lasting approximately 1 hour each Combined an empirically supported fall prevention exercise programme, a home safety evaluation, and exposure‐based CBT (manual available upon request from the first author). The CBT component included psychoeducation about anxiety and the role of avoidance, creation of a fear hierarchy based on identified triggers and avoidance behaviours, exposure practice, cognitive restructuring, and problem‐solving. The CBT component was primarily delivered in weeks 5–7, with some elements (e.g. psychoeducation about fear and avoidance, development of a fear hierarchy, relapse prevention) integrated into sessions 2, 4, and 8, respectively. Participants were offered the opportunity to invite caregivers or other support people to attend the sessions. Exercise component based on the Otago Exercise Program. Participants were additionally instructed to practice the exercises 3 times a week and walk for up to 30 minutes twice a week. Additionally, participants were taught how to get up off the floor after a fall. The home safety assessment was conducted by a therapist in session 3 based on recommendations for eliminating hazards in the home. During this session, the therapist helped participants identify and develop an action plan for addressing hazards in the home (e.g. removing throw rugs, installing grab bars, replacing dim light bulbs with brighter ones). Arm 2: control; FPE: sessions included education about reducing personal and environmental risk factors and reducing injury from falls. The education was designed to be an attention placebo condition. Delivered by physiotherapists receiving weekly supervision from a clinical psychologist. Control sessions were delivered by 1 doctoral‐level psychologist and 3 graduate students in clinical psychology who received weekly supervision.
Outcomes	FES‐I; Activity Card Sort; CIRS‐G; falls (falls diary); anonymous survey completed at home
Notes	Start and end dates not specified. No conflicts of interest declared. Source of funding: NIMH R34 MH086668. Details about the control were obtained directly from the lead author, specifying that the education was designed to be an attention placebo condition.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation using a computer‐generated sequence.
Allocation concealment (selection bias)	Low risk	Randomisation was held by a colleague with no other connection to the study.
Blinding of participants and personnel (performance bias) All outcomes	High risk	No blinding of participants.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Assessors blinded, but outcomes were self‐reported by participants who knew their group allocation.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Outcome data based on > 80% of intervention group.
Selective reporting (reporting bias)	Unclear risk	Protocol not published.
Other bias	Unclear risk	Study reported mean and SD score only in figures.

Zijlstra 2009.

Study characteristics
Methods	RCT Randomisation by an independent researcher blinded to participant characteristics performed block randomisation (2 communities × 5 cycles) using a computer‐generated random allocation.
Participants	Country: the Netherlands n = 540 (intervention: 280, control: 260) Female: intervention: 198 (70.7%), control: 190 (73.1%) Mean age (years): intervention: 77.8 (SD 4.6), control: 78.0 (SD 5.0) Inclusion criteria: community dwelling, aged ≥ 70 years, with at least some FoF and at least some activity avoidance due to FoF Exclusion criteria: confined to bed, restricted by permanent use of wheelchair, waiting for nursing home admission, or participating in other intervention
Interventions	Arm 1. AMB‐NL, a translated and adapted version of AMB; 8 × weekly sessions of 2 hours with booster session 6 months after the 8th session. Each session had predefined theme; introduction to programme, exploring thoughts and concerns about falling, exercise and fall prevention, assertiveness and fall prevention, managing concerns about falling, recognising 'fall‐ty' habits, recognising fall hazards in the home and community, practising no 'fall‐ty' habits, booster session Each session had predefined structure; introduction, participant's point of view, positive and negative aspects concerning the topic, association with falls or fear of falling, implementation in the participant's daily life Delivered by geriatric nurses working for home care agencies Arm 2: usual care
Outcomes	"Are you afraid of falling?"; FES with 4 added items about outdoor activities; PCOF; "Do you avoid certain activities due to fear of falling" (1–5 scale); FAI; MOS‐SF, item 1 (perceived general health); self‐rated life satisfaction (7 point); ADL subscale of GARS; HADS; SSL12‐I; feelings of loneliness (6‐point Likert scale); CoF; during the past 4 weeks, how often did you feel lonely (6‐point Likert scale); falls (fall calendar and 1‐item question, if medical care was provided following the fall)
Notes	Overall trial start date: 1 March 2002, overall trial end date: 31 March 2007 No conflicts of interest declared. Sources of funding: ZonMw – The Netherlands Organization for Health Research and Development (grant 014‐91‐052), CAPHRI‐School for Public Health and Primary Care, and the Faculty of Health, Medicine and Life Sciences of the Maastricht University. Protocol published
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	An independent researcher blinded to participant characteristics performed block randomisation (2 communities × 5 cycles) using a computer‐generated random allocation.
Allocation concealment (selection bias)	Low risk	Randomisation was performed directly after baseline measurements were obtained.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Blinding not possible due to nature of intervention.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Assessors blinded, but outcomes were self‐reported by participants who knew their group allocation.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Missing data were imputed.
Selective reporting (reporting bias)	Low risk	Protocol published and followed. Of note, although the study started recruitment in 2002, the overall trial end date was in 2007. The protocol was published in 2005, and is unlikely to introduce a risk of bias as it precedes the end of the trial.
Other bias	Low risk	Appeared free of other sources of bias.

ABC: Activities‐specific Balance Confidence Scale; ABLE: Activity, Balance, Learning, and Exposure intervention; ADL: activities of daily living; AMB: "A Matter of Balance"; CBT: cognitive behavioural therapy; CIRS‐G: Cumulative Illness Rating Scale for Geriatrics; COF: Consequences of Falling Scale; CST: Chair Stand Test; DSST: Digit Symbol Substitution Test; EQ‐5D: EuroQoL‐5 Dimension; EQ‐5D VAS: EuroQoL‐5 Dimensions Visual Analogue Scale; FAI: Frenchay Activities Index; FES‐I: Falls Efficacy Scale – International; FES‐IAB: Falls Efficacy Scale – International Avoidance Behavior; FES‐S: Falls Efficacy Scale, Swedish version; FoF: fear of falling; FPE: Fall Prevention Education; GARS: Gaze Anxiety Rating Scale; GDS: Geriatric Depression Scale; GFFM: Geriatric Fear of Falling Measure; HADS: Hospital Anxiety and Depression Scale; iADL: instrumental activities of daily living; ISSB: Inventory of Social Supportive Behaviors; LSNS‐6: Lubben Social Network Scale; MI: motivational interviewing; MMSE: Mini‐Mental State Examination; MOS‐SF: Medical Outcome Study Short Form; n: number of participants; PCOF: Perceived Control over Falling; OEE: Outcome Expectations for Exercise scale; OEG: Otago Exercise Group; OEP: Otago Exercise Programme; RCT: randomised controlled trial; SD: standard deviation; SEE: Self‐efficacy for Exercise scale; SF‐12: 12‐Item Short‐Form Health Survey; SF‐36: 36‐item Short Form Health Survey; SIP: Sickness Impact Profile; SPMSQ: Short Portable Mental Status Questionnaire; SPPB: Short Physical Performance Battery; SSL12‐I ‐ Social Support List of interactions Interaction version; TUG: Timed Up and Go Test; WHOQOL‐BREF: World Health Organization Quality Of Life questionnaire Brief Version; YPAS: Yale Physical Activity Survey.

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Azizan 2015	Ineligible intervention: the behavioural component did not fulfil the requirements to be considered CBT.
Banez 2008	Ineligible study design: this was a pre–post intervention study. Furthermore, the interprofessional team did not consist of elements that would be considered CBT.
Brouwer 2003	Ineligible intervention: education component did not fulfil requirements to be CBT.
Chen 2014	Ineligible study design: a doctoral dissertation that included studies that did not fulfil our inclusion criteria for study design.
Dattilo 2014	Ineligible intervention: the education used did not fulfil the criteria to be considered CBT.
Duenas 2019	Ineligible control group: included 125 participants randomised to Tai Chi, CBT, and postural control exercise groups. The control group was an exercise intervention that did not fulfil the criteria for inclusion in our review.
Gill 2020	Ineligible intervention: the multifactorial intervention consisted of assessment, recommendations, motivational interviewing, developing individualised care plans, and implementing the fall care plans. The motivational interviewing component did not fulfil the requirements to be considered CBT.
Headley 2014	Ineligible intervention: the N'Balance programme is an exercise intervention, and does not have a component that would fulfil the criteria for CBT interventions.
IRCT20211201053248N1	Ineligible control group: control group consisted of an exercise intervention; "conventional gait rehabilitation."
Johansson 2018	Ineligible intervention: intervention programme used a small‐group learning environment and peer learning, with an occupational science focus. It did not fulfil the criteria of CBT.
Kwon 2011	Ineligible intervention: study investigated an exercise intervention.
Lee 2013	Ineligible intervention: intervention group received a risk‐based multifactorial fall prevention programme including exercise intervention, health education, and home hazards evaluation/modification, along with medication review and referral to an ophthalmologist or other specialists.
Lim 2023	Ineligible intervention: online self‐guided intervention that did not fulfil the studies as the therapy was not provided by a person, and as such lacked the interactive nature required for CBT
Lin 2007	Ineligible intervention: this 3‐armed study had an exercise intervention, home safety assessment group, and educational group. There was no CBT component.
Liu 2014	Ineligible control: study compared a CBT group with Tai Chi to a Tai Chi‐only group. As the control group received exercise, it did not fulfil the eligibility criteria for this review.
NCT01268657	Ineligible study design: study delivered CBT intervention. However, it had only 1 arm, and was not a randomised controlled trial.
NCT02727374	Ineligible control group: study investigated a physiotherapy intervention with CBT intervention. The control group was a physiotherapy intervention.
NCT03211429	Ineligible control group: 3‐armed randomised controlled group, based in Manizales (Colombia), that investigated a CBT intervention for reducing fear of falling, with Tai Chi, and Postural Control exercises, and did not fulfil the inclusion criteria of this review.
Sartor‐Glittenberg 2018	Ineligible study design: quasi‐experimental, 1‐group, pretest–post‐test study design investigated the avoidance outcomes from the A Matter of Balance intervention.
Suttanon 2018	Ineligible intervention: exercise intervention based on the Otago programme. No CBT component.
Thiamwong 2019	Ineligible study design: the Physio‐fEedback Exercise pRogram (PEER) intervention consists of visual physio‐feedback on balance, cognitive reframing, and peer coaching. However, the study did not have a control group, and did not fulfil the requirements for inclusion in this review.
van Schooten 2021	Ineligible intervention: the behavioural component did not fulfil the requirements to be considered CBT – the online component lacked the interactivity of this review's definition of CBT.
Walters 2018	Mismatch with study design: a quantitative descriptive study of an "A Matter of Balance" intervention that had been in place at the senior centre, as part of its regular programming.
Wolfe 2018	Mismatch with study design: study evaluated the change in fear of falling in older adults enrolled in either Chinese‐ or Spanish‐speaking "Matter of Balance" programmes. There was no non‐interventional control arm.

CBT: cognitive behavioural therapy.

Characteristics of ongoing studies [ordered by study ID]

ACTRN12621000440820.

Study name	Own Your Balance Study: effect of a self‐managed online cognitive behavioural therapy program in older people with concerns about falling
Methods	3‐arm RCT
Participants	Inclusion criteria: aged ≥ 65 years, concerned about falls or have low balance confidence (or both), living in the community, English speaking, independent in activities of daily living, able to walk household distances with or without the use of a walking aid, willing to give informed consent and comply with the study protocol Exclusion criteria: cognitive impairment, severe depression or suicidal thoughts, acute psychiatric condition with psychosis, any medical condition that precludes exercise participation, progressive neurological conditions (such as Parkinson's disease, multiple sclerosis, Alzheimer's disease), currently participating in a fall prevention programme
Interventions	2 intervention groups (Own Your Balance Program and Own Your Balance Program plus the StandingTall graded activity program) delivered via tablet or computer in people's homes with limited therapist input over 8 weeks. After 8 weeks, both groups will receive individual guidance by telephone, using motivational interviewing techniques, to use the Active and Healthy website to continue with an exercise programme (≥ 2 hours of exercise each week) for the next 10 months. A follow‐up call will be performed at 12 weeks to see if the participants in both intervention groups have started with their exercise programme.
Outcomes	n = 189 Primary outcome: IconFES Secondary outcomes: IconSAFE; Attitudes to Falls‐Related Interventions Scale; DASS‐21, depression, anxiety and stress subscale; Health services used (monthly health services access questionnaire); PGCI; EQ‐5D; PACES; ABC; IconFES; FHLS; IPEQ; rate of falls (self‐reported); MOS‐SS; System Usability Scale; overall movement intensity (McRoberts MoveMonitor activity monitor); COMPAS‐W scale; adherence to exercise programmes (self‐reported); adherence to the intervention; ESES; habitual daily activity (McRoberts MoveMonitor activity monitor)
Starting date	7 June 2021
Contact information	Professor Kim Delbaere; Neuroscience Research Australia Margarete Ainsworth Building Barker Street Randwick Sydney NSW 2031 Australia; +61 2 93991066; k.delbaere@neura.edu.au
Notes	Completion of data collection expected by July 2023.

NCT05192408.

Study name	Multi‐component Intervention for reducing fear of falling in community‐dwelling older adults
Methods	RCT
Participants	n = 420 Inclusion criteria: adults aged ≥ 65 years; followed up at National Healthcare Group Polyclinics for any chronic condition; able to communicate in English, Mandarin, and Malay; and reports fear of falling as screened by single question. Exclusion criteria: unable to agree to participation, unable or unsafe to participate in any exercise as determined by clinician (e.g. terminal illness, uncontrolled hypertension, immobility, unable to participate in telephone sessions e.g. due to hearing impairment), and known psychiatric condition on treatment
Interventions	Intervention (individual): 4 sessions face‐to‐face or via videoconferencing. CBT component comprises behavioural activation, cognitive restructuring, promotion of safety, relaxation techniques, and motivational interviewing for goal setting. Exercise component comprises patient education materials and videos tailored to frailty level and readiness to progress, with balance and strength components. Control: patient education materials on: exercise in older adults (Stay Active, Stay Strong and Stay Steady), falls prevention, and fear of falling.
Outcomes	Primary outcome: fear of falling measured using short FES‐I Secondary outcome: number of falls, physical activity measured using IPEQ‐WA
Starting date	3 January 2022
Contact information	Jacqueline G De Roza, National Healthcare Group Polyclinics. Singapore, Singapore, 380130. +6594500212, Jacqueline_G_De_ROZA@nhgp.com.sg
Notes	Expected to be completed May 2023

Taylor 2021.

Study name	Motivational interviewing to increase walking in community‐dwelling older adults after hip fracture
Methods	RCT
Participants	n = 270 Inclusion criteria: had a hip fracture, live at home independently, aged ≥ 65 years, have been discharged from hospital within the last 6 months, able to walk independently with or without an assistive device, can communicate with conversational English, insufficiently active (defined as obtaining < 150 minutes of moderate‐intensity activity per week) Exclusion criteria: severe depression or anxiety, score > 2 errors on the 10‐item Short Portable Mental Status Questionnaire indicative of impaired intellectual functioning, medically unstable to walk, unable to converse by telephone, live in residential care
Interventions	Intervention: telephone‐based motivational interviewing to collaboratively work with participants to assist them to increase their motivation to walk Control: standard care
Outcomes	Primary outcome: daily time spent walking (accelerometer‐based activity monitor (activPAL)) Secondary outcomes: 22‐item Ambulatory Self‐Confidence Questionnaire; DASS‐21; feasibility, advantages and disadvantages of the intervention (inductive thematic analysis of therapists); perceptions about receiving the motivational interviewing (inductive thematic analysis of participants); injuries as a result of a fall (self‐reported); daily steps (accelerometer‐based activity monitor (activPAL)); participant‐report of whether they walk outdoors alone or with company, and to report the frequency of outdoor walks (self‐reported); falls (self‐reported); daily time spent in moderate‐to‐vigorous activity (activPAL); cost‐effectiveness from a healthcare system perspective; hospital readmissions; daily time sitting or lying (activPAL); nutritional status (Malnutrition Screening Tool); bodyweight; well‐being and capability (ICECAP‐O); medical services and pharmaceutical use (Medicare Australia records); FAI; exploration of experiences of rehabilitation after hip fracture, and recovery of mobility for a purposive sample of participants in both groups (inductive thematic analysis of participants); AQoL; healthcare utilisation and cost (patient health service utilisation questionnaire developed for the trial); m‐FES
Starting date	29 September 2019
Contact information	Professor Nicholas Taylor; Allied Health Clinical Research Office (Eastern Health / La Trobe University) level 2/5 Arnold Street Box Hill VIC 3128 Australia; +61 3 90918874; n.taylor@latrobe.edu.au
Notes	Recruitment phase completed. Expected results in 2023.

ABC: Activities‐specific Balance Confidence scale; AQoL: Assessment of Quality of Life Instrument; DASS‐21: 21‐item Depression, Anxiety and Stress Scale; EQ‐5D: EuroQoL‐5 Dimension; ESES: Exercise Self Efficacy Scale; FAI: Frenchay Activities Index; FHLS: Fall‐Related Health Literacy; ICECAP‐O: ICEpop CAPability measure for Older people; IconFES: Iconographical Falls Efficacy Scale; IconSAFE: Iconographical Survey of Activities and Fear of Falling in the Elderly; IPEQ: Incidental and PlannEd activity Questionnaire; IPEQ‐WA: Incidental and Planned Exercise Questionnaire for the Usual Week; PGCI: Patient Global Impression of Change Scale; m‐FES: modified Falls Efficacy Scale; MOS‐SS: Medical Outcomes Study Social Support Survey; n: number of participants; PACES: Physical Activity Enjoyment Scale; RCT: randomised controlled trial.

Differences between protocol and review

We made the following changes from the published protocol (Lenouvel 2021b).

• Title: we changed the title from "Cognitive behavioural interventions for reducing fear of falling in older people living in the community" to "Cognitive behavioural therapy (CBT) with and without exercise to reduce fear of falling in older people living in the community." Cognitive behavioural therapy is commonly referred to as only 'CBT'. The addition of CBT in the title was to add clarity and ease of identifying this article when searching CBT. The addition of 'with and without exercise' is to better reflect the objective of the review: to assess the effects of CBT only and CBT with exercise.

• Introduction: we modified the text to improve readability.

• Methods: we reformulated the text to improve readability.

  • We removed the definitions of cluster and quasi‐experiment studies, as we consider this to be superfluous.

  • We specified the definition of comparative studies as "… studies that did not include CBT with or without exercise interventions." This clarifies the inclusion of relevant articles, such as Huang 2011, which is a three‐arm study, comparing CBT alone, and CBT with exercise, with a control group.

  • Types of interventions: we removed the SAFFE questionnaire as a potential primary outcome measure of FoF, as it appeared in both primary and secondary outcomes. We added the FES‐IAB to possible secondary outcome measures for activity avoidance.

  • We considered only English and German language articles for inclusion due to language limitations of the review team.

  • We specified how studies with multiple arms would be analysed in the quantitative synthesis.

  • Sensitivity analysis of missing data was not performed (removing quasi‐RCTs).

  • Unit of analysis issues; we specified how unit of analysis issues would be accounted for and mentioned how we would incorporate studies with multiple arms that were not cross‐over studies.

  • We removed sustainability of effects as a secondary outcomes. This was due to unanticipated generalised insufficiency of data, due to a low number of studies reporting the secondary outcomes, and a low number of participants. We did not consider that reporting of sustainability of effects of secondary outcomes added value to this review.

  • Measurement of effect; we specified how sustainability of effect was classified as less than six months and six months or greater after the end of treatment. We considered this categorisation would better demonstrate the effects of FoF interventions over time, taking into account the variations in measurements between studies.

  • We removed "and short‐ and long‐term measures of well‐being and QoL" from the secondary analyses, as this was a repetition from one of the prementioned secondary outcomes (QoL).

  • Table 1: following the reasoning mentioned above, we added sustainability of effects, more than six months' postintervention. We did not include anxiety outcomes, as we considered this secondary outcome was less relevant, particularly as the review did not yield significant data.

  • We removed the characterisation of exercise components using the ProFaNE taxonomy, as we considered that the focus of this review was on the CBT interventions rather than the exercise components, and as such beyond the scope of this review.

  • We added an explanation of how we calculated the change score, as it was missing in the original protocol.

  • We removed the mention of searching the Cochrane Bone, Joint and Muscle Trauma (BJMT) Group Specialised Register, as Cochrane CENTRAL included the Cochrane BJMT group register.

• Analysis

  • Subgroup analysis

    • Due to insufficient data in included studies, baseline fall risk, cognitive state, and sex could not be included for analysis. Baseline FoF could not be included due to difficulties in comparing the different FoF measures.

    • We presented in the analyses CBT‐only interventions and CBT with exercise interventions subgroups. We considered this clinically meaningful, illustrating the effects of these two major classes of interventions.

  • Sensitivity analysis

    • We conducted additional sensitivity analyses post hoc in trials for primary outcomes where: 1. FoF was not the primary outcome; 2. studies were clustered (due to potential ICC issues); 3. when none‐FES based questionnaires were used as measures, and 4. for trials where active control groups were used. We included these analyses as we considered it would improve the quality of the review. As there were no quasi‐RCTs in this study, they could not undergo sensitivity analysis, and as such were not included in the list of sensitivity analyses.

• Outcomes for presentation in Table 1

  • We added the outcome sustainability of effects more than six months, as we considered this more clinically relevant to the long‐term benefits of CBT interventions for reducing FoF. We removed anxiety from the secondary outcomes listed in the Table 1.

Contributions of authors

EL: lead investigator, searched literature, checked reference lists for additional studies, database management, data extraction, analysis, writing manuscript, revision of manuscript. EL is the guarantor of the review.

PU: searched literature, checked reference lists for additional studies, database management, data extraction, writing manuscript, revision of manuscript

WS: provided methodological expertise and statistical advice, revision of manuscript

DD: contributed to drafting the protocol, provided statistical advice, revision of manuscript

MD: contributed to drafting the protocol, revision of manuscript

GK: contributed to drafting the protocol, adjudicated on papers for inclusion, risk of bias assessment, revision of manuscript

GARZ: contributed to drafting the protocol, revision of manuscript

KH: adjudicated on papers for inclusion, data extraction, data analysis, writing manuscript, revision of manuscript

SK: lead supervisor of review team, co‐ordinated protocol design, database management, data extraction, analysis, writing manuscript, revision of manuscript

Sources of support

Internal sources

• Department of Old Age Psychiatry and Psychotherapy, University Hospital of Psychiatry, Bern, Switzerland

  Salary support for EL, L Novak (contributor to the protocol), T Wirth (contributor to the protocol), SK.

• University of Bern, Graduate School for Health Sciences (GHS), Other

  Library resources

• AGAPLESION Bethesda Clinic Ulm, Ulm, Germany, Other

  Salary support MD, DD

• Institute for Evidence in Medicine (for Cochrane Germany Foundation), Medical Center – University of Freiburg, Freiburg, Germany, Other

  Salary support SV, library resources

• Geriatric Research Institute, University of Ulm, Ulm, Germany, Other

  Library resources

External sources

• None, Other

  No specific grants from any funding agency in the public, commercial, or non‐profit sectors were received.

Declarations of interest

EL: was granted funding from the Age Stifung (Kirchgasse 42, 8001 Zürich, Switzerland) during the process of this review for developing a cognitive behavioural therapy (CBT) intervention for reducing fear of falling in nursing home residents, which when the study is conducted, due to a different population would not fulfil the inclusion criteria of this review.

PU: none.

WS: none.

DD: none.

MD: none.

GK: none.

GARZ: was involved in conducting a study that was included in the review (Zijlstra 2009). She was independent of the study selection decision, risk of bias assessment, and data extraction for this study.

KH: none.

SK: was granted funding from the Age Stifung (Kirchgasse 42, 8001 Zürich, Switzerland) during the process of this review for developing a CBT intervention for reducing fear of falling in nursing home residents, which when the study is conducted, due to a different population would not fulfil the inclusion criteria of this review.

New