Associations between perfectionism and symptoms of anxiety, obsessive-compulsive disorder and depression in young people: a meta-analysis

ABSTRACT

Perfectionism is a transdiagnostic process contributing to the onset and maintenance of anxiety, obsessive-compulsive disorder (OCD) and depression. The aim of this systematic review and meta-analysis was to examine the association between perfectionism, and symptoms of anxiety, OCD and depression among young people aged 6–24 years. A systematic literature search retrieved a total of 4,927 articles, with 121 studies included (M_{pooled age} = ~17.70 years). Perfectionistic concerns demonstrated significant moderate pooled correlations with symptoms of anxiety (r = .37–.41), OCD (r = .42), and depression (r = .40). Perfectionistic strivings demonstrated significant, small correlations with symptoms of anxiety (r = .05) and OCD (r = .19). The findings highlight the substantial link between perfectionistic concerns and psychopathology in young people, and to a smaller extent perfectionistic strivings, anxiety, and OCD. The results indicate the importance of further research on early intervention for perfectionism to improve youth mental health.

KEYWORDS:

• Perfectionism
• anxiety
• depression
• obsessive-compulsive disorder
• young people
• meta-analysis

View correction statement:

Correction

Perfectionism is a transdiagnostic process, contributing to the aetiology and maintenance of anxiety, obsessive-compulsive disorder (OCD) and depression (Egan et al., 2011). Perfectionism has predominately been measured with the Multidimensional Perfectionism Scales (FMPS; Frost et al., 1990; HMPS; Hewitt & Flett, 1991). The FMPS (Frost et al., 1990) consists of subscales measuring personal standards, concern over mistakes, doubts about actions, parental expectations, and parental criticism. The HMPS (Hewitt & Flett, 1991) consists of three subscales: self-oriented perfectionism: setting high personal standards; socially prescribed perfectionism: belief that others expect high standards for the individual and other-oriented perfectionism: belief that others should be perfect. Factor analyses of the scales result in a consistent two-factor structure of perfectionistic strivings (personal standards, self-oriented and other-oriented perfectionism) and perfectionistic concerns (concern over mistakes, doubts about actions, parental expectations and criticism and socially prescribed perfectionism) (Stoeber & Otto, 2006).

Another definition of perfectionism is clinical perfectionism, where self-esteem is based on striving for high standards despite negative effects, including symptoms of anxiety, depression and OCD (Shafran et al., 2002). It is hypothesised in the cognitive-behavioural model that self-worth based on achievement leads to symptoms of a range of psychological disorders as the individual is continually striving for high standards and engaging in self-criticism regardless of whether personal standards are met (Shafran et al., 2002). There have been various cross-sectional and experimental studies that have provided support for the cognitive-behavioural model of clinical perfectionism (see Shafran et al., 2023 for a review). Cognitive-behaviour therapy for perfectionism (CBT-P) is based on the cognitive-behavioural model and has been described as the leading evidence-based treatment for perfectionism (Egan, Shafran, et al., 2022). CBT-P has demonstrated efficacy in the reduction of symptoms of anxiety and depression in over 15 randomised controlled trials in a recent meta-analysis of studies of CBT-P predominately in adults (Galloway et al., 2022). Some authors in defining perfectionism have emphasised the distinction between “healthy” or “adaptive” perfectionism, for example, based on studies finding an association between perfectionistic strivings and positive outcomes (e.g. Stoeber et al., 2020). Others have argued there is a need to separate the pursuit of excellence, termed “excellencism” from perfectionism, arguing that over and above excellencism, perfectionism is not associated with adaptive outcomes (Gaudreau et al., 2022).

There have been many studies that have examined the association between perfectionism and psychopathology. Limburg et al. (2017) performed a meta-analysis of this literature in adult samples and found both perfectionistic strivings and perfectionistic concerns had significant associations with psychopathology Perfectionistic concerns had higher associations with symptoms of anxiety (r = .35), OCD (r = .30), and depression (r = .39) than perfectionistic strivings had with symptoms of anxiety (r = .14), OCD (r = .14), and depression (r = .11) (Limburg et al., 2017). However, Limburg et al. (2017) excluded participants under the age of 18 years therefore generalisations cannot be made to younger people. There has been no meta-analysis to date that has examined the association between perfectionism and psychopathology specifically among young people.

Smith, Sherry, et al. (2018) conducted a meta-analysis including 11 longitudinal studies on the association between perfectionism and symptoms of anxiety, in samples including children, adolescents and adults. There were significant associations between symptoms of anxiety and the different perfectionism subscales, which assess perfectionistic concerns, such as concern over mistakes (r = .11) and doubts about actions (r = .13), in line with Limburg et al. (2017). However, Smith, Sherry, et al. (2018) found inconsistent results with perfectionistic strivings, with non-significant associations between self-oriented perfectionism and symptoms of anxiety (r = .03), but significant, albeit small, associations between personal standards and symptoms of anxiety (r = .08).

There is a need for a meta-analysis to examine the relationships between perfectionistic strivings, concerns, and psychopathology in young people, given debate and disagreement in the literature whether strivings have a positive association with psychopathology (see Limburg et al., 2017) or not, and whether age has an influence on the relationship between strivings and concerns. Although Smith, Sherry, et al. (2018) included young people in their meta-analysis, it was not disaggregated by age, which is problematic since pooling associations across children and adults does not clearly inform the relationship in youth. Conducting a meta-analysis on the relationship between perfectionism and symptoms of anxiety, depression and OCD in young people is important to assist in informing treatment, prevention and support strategies for young people with elevated perfectionism (Egan, Wade, et al., 2022).

Young people should be considered as a specific population as given the developmental differences across age ranges, it cannot be assumed that psychological theories and treatments that apply to adults directly apply to youth (Bastien et al., 2020). Studies examining the relationship between age and perfectionism have demonstrated mixed results, some have shown a positive relationship between total perfectionism and age (e.g. Butt, 2010), while others have found no significant relationship with age (e.g. Schweitzer & Hamilton, 2002). Sand et al. (2021) argued that age is an important variable to examine in relation to perfectionism given adolescents and young people are at a sensitive point for increases in perfectionism due to awareness of striving to achieve standards and self-consciousness. Sand et al. (2021) conducted a large sample with 10,217 adolescents aged 16–19 years from Norway and found no difference in levels of perfectionism over this age range. However, it would be useful to examine a wider age range of youth in a meta-analysis pooling the relationship between perfectionism dimensions and psychopathology across studies to inform understanding of if there are any age effects.

In this meta-analysis, we defined a young person as between 6 and 24 years. The lower age of 6 years was based on a preliminary search of the lowest age range of studies on the association between perfectionism and psychopathology. The upper age range of 24 years is in accordance with the World Health Organization’s (2014) definition of a young person, as many consider adolescence as extending into the early twenties, including based on evidence that brain maturation does not occur until the mid-20s (Steinberg, 2014). Symptoms of anxiety, depression and OCD were examined given they are the most prevalent and commonly seen in clinical practice. Depression has a prevalence rate of 12.9% in adolescents and a high chance of continuing into adulthood, and anxiety is the psychological disorder with the highest prevalence among young people, largely continuing to adulthood (Polanczyk et al., 2015). OCD affects 1–2% of children, with 50% of adults diagnosed with OCD reporting the onset of symptoms before 18 years of age (Averna et al., 2018). Despite symptoms of eating disorders being a focus of a review in adults (Limburg et al., 2017), it was excluded in this meta-analysis due to being examined in a concurrent separate review in young people (Prospero CRD42022324655). The reason for separating out symptoms of eating disorders is that the treatment of eating disorders is distinct from anxiety, depression and OCD, for example, the treatment of choice for eating disorders in adolescents is family-based treatment (Lock & Le Grange, 2015) versus CBT for symptoms of anxiety, depression and OCD in youth (Pote, 2021).

The aim of the study was to perform a meta-analysis on the associations between perfectionism and symptoms of anxiety, OCD and depression in young people aged 6–24 years. We hypothesised that there would be a medium pooled correlation between perfectionistic concerns and symptoms of anxiety, OCD and depression, and a small pooled correlation between perfectionistic strivings and symptoms of anxiety, OCD, and depression. To examine the potential role of moderators, age, clinical versus a non-clinical sample, and high income versus low-middle-income country (LMIC) of residence were examined. It was predicted relationships between perfectionism dimensions and symptoms would be stronger in clinical versus non-clinical samples based on a previous meta-analysis in adults (Limburg et al., 2017). No specific hypotheses were made for age and high versus LMIC country of residence, due to these moderating factors not previously being included in the meta-analysis in adults on the association between perfectionism and psychopathology (Limburg et al., 2017).

Method

Inclusion and exclusion criteria

The inclusion criteria were: (i) in English, (ii) include a validated measure of symptoms of anxiety, and/or OCD, and/or depression, (iii) include a validated self-report measure of perfectionism which can be classified into strivings and concerns, (iv) include participants aged between 6 and 24 years, if there is no age range provided a mean age between 6 and 24 years, (v) report a correlation between perfectionistic strivings and/or concerns, with symptoms of either anxiety, OCD, and/or depression, or provide results which allow for a correlation to be calculated, and (vi) peer-reviewed. Exclusion criteria were: (i) if a correlation cannot be computed from the results, (ii) qualitative design, (iii) a dissertation, (iv) measured performance and/or sports anxiety or perfectionism, (v) used a sample with clinical or subclinical eating disorder symptoms and/or an eating disorder diagnosis, (vi) measures of perfectionism, and symptoms of anxiety, OCD and/or depression with a Cronbach’s alpha <.70.

Databases and search strategy

PRISMA guidelines were followed (Page et al., 2021). The search strategy was registered on PROSPERO on 21 April 2022, prior to the search (CRD42022327290). Databases were searched on 22 May 2022, including Medline, Scopus, Proquest and PsycINFO. The search terms were (perfectionis*) AND (depress* OR anxiet* OR psychopatholog* OR “obsessive compulsive” OR “mental health” OR OCD) AND (child* OR adolescent* OR teen* OR youth OR “young adult” OR college OR undergraduate OR student). There were no restrictions placed on publication date.

Study identification and inter-rater reliability

The study selection process was carried out by JL. If studies were missing specific age information, authors were contacted. If age could not be retrieved, the study was excluded. A secondary reviewer (TC) independently screened 30% of articles at title/abstract and full-text level. Cohen’s kappa was used for inter-rater reliability; poor (k < 0.00), slight (k = 0.00–0.20), fair (k = 0.21–0.40), moderate (k = 0.41–0.60), substantial (k = 0.61–0.80) or almost perfect (k = 0.81–1.00) (Landis & Koch, 1977).

Data extraction and management

Division of perfectionism measures into perfectionistic strivings and concerns was based on Limburg et al. (2017) (see table in supplementary materials). Age was coded based on mean age into (0) 6–12.99 years (1) 13–24 years. The rationale for this age split was to separate samples into children and adolescents/young adults, to inform future treatment research, given treatments differ between these two age ranges. For example, treatment is generally parent-led in children aged 6–12.99 years, while for young people aged 13–24 treatment is typically engaged in individually by the young person rather than led by their parents. Anxiety symptoms were categorised into a general category of “anxiety”, referring to anxiety symptoms including, for example, generalised anxiety and social anxiety, largely being based on anxiety symptoms of the anxiety disorders listed in the Diagnostic and Statistical Manual of Mental Disorders (5th ed.; DSM) criteria (American Psychiatric Association, 2013). Test anxiety symptoms were considered as a separate category, which fit under the definition of test anxiety, where extreme distress is experienced in test-taking environments (Abdollahi & Abu Talib, 2015), as this was a specific literature in relation to association with this form of anxiety symptoms and perfectionism in young people.

Assessment of heterogeneity

Heterogeneity was assessed using the I² and Cochran’s Q statistics. Cochran’s Q is calculated using the total squared deviations of the effect sizes from each study, with a significant p-value demonstrating heterogeneity. The I² shows the percentage of variation which is caused by heterogeneity, with 25% being low, 50% being moderate and 75% being high heterogeneity. When I² is 60% or greater potential moderators should be considered in a meta-regression to identify sources of heterogeneity (Baker et al., 2009).

Quality assessment

The quality of the studies was assessed by JL using the 14 item National Institute of Health (NIH) quality assessment tool for observational and cross-sectional studies (National Institute of Health, 2021). The items including quality of the study design, confounding variables, measurement bias and information bias, are rated on a “0 = no” or “1 = yes” scale (see quality rating table in supplementary materials). The quality score was derived by dividing the total number of “yes” responses by the number of items, with scores multiplied by 100 to produce a percentage, where less than 50% = poor, 51–74.99% = fair, and greater than 75% = good.

Publication bias

Publication bias was determined using Egger’s test for plot asymmetry. There was also visual inspections of the funnel plots for asymmetry (Egger et al., 1997).

Data analysis

Pearson’s correlation coefficient (r) was used to determine the associations between perfectionism and psychopathology symptoms. For between groups studies, effect size was calculated using the means and standard deviations between the treatment/clinical group and the control/non-clinical group. In longitudinal studies, only the correlations for the first-time point were imputed. If subscale scores were provided where multiple subscales fit in perfectionistic strivings or concerns, the individual correlations for the subscales were combined and averaged to determine the overall strivings or concerns correlation.

A meta-analysis was conducted to determine the associations between perfectionism and symptoms of anxiety, OCD and depression in JASP (JASP Team, 2018). A Hunter-Schmidt random effects model was used to determine the pooled random effects for the studies. Effect sizes were determined using Cohen’s (1992) principles, where small (r = .10), medium (r = .30), and large (r = .50). The 95% confidence intervals were also calculated, mean of the distribution (Z) and standard error (SE). The moderators of age (categorised as 6–12.99 and 13–24) and country of residence (categorised as high income or LMIC) were explored in meta-regressions. We examined potential influential outlying cases using case diagnostics tools in JASP. Cases are identified as influential by Cooks’ distance with the general rule of thumb of any case with a Cook’s distance larger than 4/n (data points) being considered influential (Viechtbauer & Cheung, 2010). For associations that had influential cases, we ran pooled effects with and without these cases and comment on any substantial differences in pooled effects.

Results

Study identification and inter-rater reliability

There were 4,927 studies identified, with 2581 duplicates removed (see Figure 1). This left 2,346 articles screened at title/abstract by JL, with 30% independently screened by TC. JL and TC were trained by SE in screening and judgement of inclusion and exclusion criteria. There was substantial interrater reliability at title/abstract screening (k = .61). There were 1,478 articles excluded at title/abstract level, resulting in 868 articles screened at full-text level by JL. In articles where the age range and mean age were unclear, authors were contacted. A total of four authors were contacted for six articles, but only one responded. The articles where there was no response were excluded. Only one article was unable to be retrieved. TC screened 30% of articles at full text, with substantial interrater reliability (k = .70). There were 121 eligible articles. Any articles where there was a discrepancy were read at full text by the senior author (SE) until consensus was reached.

Figure 1. Prisma 2020 flow diagram of the screening process.

Note: n = number of studies.

Study characteristics

There were 121 articles which met the inclusion criteria (see Table 1 for study characteristics). Studies included a total of 41,824 participants (M _{pooled age} = ~17.70 years, SD = 4.19, range = 7–24 years). Based on mean age or the midpoint of an age range if mean age was not stated, there were 17 (14%) articles with a sample aged 6–12.99 years, and 91 (76%) articles with a sample aged 13–24 years. The studies were predominately from high-income (HI) countries (n = 93; 78%). Studies were largely from the USA/Canada/Australia/UK/Europe (n = 87; 72%), specifically, USA, n = 38 Canada, n = 31¹; Europe, n = 8 (Spain, n = 2; Belgium, n = 2; Romania, Portugal, Norway, Italy all n = 1), Australia, n = 6 and the UK, n = 4. There were a small number of studies from high-income nations in the Middle East and Asia (n = 6; 5%; Israel, n = 3; South Korea, Singapore, Japan all n = 1). There were (n = 27; 21%) studies from low- and middle-income countries (LMIC) in Asia and South-East Asia (n = 14; 12%, China, n = 11,¹ Malaysia, n = 2; India n = 1), the Middle East (n = 5; 4% - Iran, n = 4; Jordan, n = 1), South America (n = 3, 2%; Ecuador, n = 2; Argentina n = 1) and Europe (n = 5, 2% - Russia, n = 2; Turkey, n = 3). For two studies, the country of residence of participants was not specified. The majority of studies measured symptoms of depression (n = 95), and/or examined symptoms of anxiety (n = 59), and/or OCD (n = 6). Most studies were cross-sectional (n = 104), with a small number of intervention (n = 9) and longitudinal (n = 8) studies (note data was taken from these studies only at baseline). Most samples were non-clinical (n = 117), with only four clinical samples.

Table 1. Study characteristics.

First author/year	Country	Design	Sample type	Sample size	Female (%)	Age range (years)	Mean age (years)	Measure of perfectionism	Measure of anxiety	Measure of depression	Measure of OCD
Abdollahi (2019)	Malaysia	Cross-sectional	University students	450	59%	18–24	19.34	APS-R	LSAS	X	X
Abdollahi and Abu Talib (2015)	Iran	Cross-sectional	High school students	520	47%	15–21	17.25	APS-R	TAS	X	X
Abdollahi et al. (2018)	Malaysia	Cross-sectional	University Students	509	55%	19–24	20.18	APS-R	X	BDI	X
Affrunti and Woodruff-Borden (2016)	USA	Cross-sectional	Children	61	46%	7–13	9.2	CAPS	BAI	BDI	X
Afshara et al. (2011)	Iran	Cross-sectional	School students	793	50%	ND	M = 14.9 F = 15	PANPS	RCMAS	CDI	X
Aldahadha and Balsamo (2018)	Jordan	Cross-sectional	University students	487	48%	ND	19.12	PANPS	DASS-21	DASS-21	X
Arana and Furlan (2016)	Argentina	Cross-sectional	University students	277	81%	ND	23.91	APS-R	GTAI	X	X
Ashby et al. (2011)	ND	Cross-sectional	Middle school students	153	61%	11–15	12.9	APS-R	X	KDI	X
Athulya et al. (2016)	India	Cross-sectional	University students	192	69%	ND	21.22	FMPS	DASS-21	DASS-21	X
Bardone-Cone et al. (2017)	USA	Cross-sectional	University students	441	100%	17–24	18.7	FMPS	STAI	X	X
Besharat et al. (2014)	Iran	Cross-sectional	University students	300	52%	ND	19.7	TMPS	RAAS	CES-D	X
Brown and Kocovski (2014)	ND	Intervention	Students	104	69%	ND	18.58	FMPS, HMPS	LSAS	BDI	X
Buzzai et al. (2020)	Italy	Cross-sectional	High school students	272	56%	ND	12.8	APS-R	MeMa	X	X
Chai et al. (2020)	China	Cross-sectional	University students	918	63%	ND	19.78	APS-R	X	CES-D	X
Chang (2013)	USA	Cross-sectional	University students	532	58%	ND	20.04	FMPS	BAI	BDI	X
Chang et al. (2011)	USA	Cross-sectional	Latino students	121	100%	ND	19.78	FMPS	BAI	BDI	X
Cheng et al. (1999)	China	Cross-sectional	Adolescents	947	59%	13–18	14.76	FMPS	DASS	DASS	X
Christopher and Shewmaker (2010)	USA	Cross-sectional	Children	240	42%	7–14	ND	CAPS	RCMAS	CDI	X
Cooks and Ciesla (2019)	USA	Intervention	University students	125	74%	ND	19	FMPS, HMPS	X	CES-D	X
Damian et al. (2017)	Romania	Longitudinal	High school students	489	54%	12–19	15.9	CAPS, FMPS	SCARED	X	X
Demirtas-Zorbaz (2020)	Turkey	Cross-sectional	High school students	640	42%	14–18	ND	PANPS	SASA	X	X
Diamond et al. (2012)	Israel	Intervention	University students	96	100%	22–24	ND	HMPS	X	CES-D	X
Dittner et al. (2011)	UK	Intervention	University students	436	71%	ND	20.4	HMPS	X	CES-D	X
Einstein et al. (2000)	Australia	Intervention	High school students	772	66%	15–24	17.6	HMPS	DASS-21	DASS-21	X
Erozkan et al. (2011)	Turkey	Cross-sectional	High school students	598	51%	ND	16	FMPS	X	BDI	X
Etherson et al. (2022)	UK	Longitudinal	University students	181	48%	ND	20.34	HMPS	X	CES-D	X
Eum and Rice (2011)	USA	Intervention	University students	134	62%	18–23	19.08	APS-R	CTS	X	X
Eum and Rice (2021)	USA	Cross-sectional	University students	213	75%	ND	20	APS-R	X	X	YBOCS
Fedewa et al. (2005)	USA	Cross-sectional	University students	230	76%	ND	19.8	PANPS	STAI	X	X
Ferrari et al. (2018)	Australia	Cross-sectional	Adolescents	541	82%	13–15	14.08	CAPS	X	SMFQ	X
Flett et al. (1991)	Canada	Cross-sectional	University students	164	62%	ND	20.99	HMPS	X	BDI	X
Flett et al. (2003)	USA	Cross-sectional	University students	94	81%	ND	21.3	HMPS	X	CES-D	X
Flett et al. (2005)	Israel	Cross-sectional	Community sample	420	26%	ND	23.5	HMPS	X	CES-D	X
Flett, Coulter, et al. (2011)	USA	Cross-sectional	Primary students	81	54%	ND	12.8	CAPS	X	CES-D	X
Flett, Panico, et al. (2011)	Canada	Cross-sectional	Adolescents	73	53%	16–18	16.8	CAPS	X	CDI	X
Flett, Druckman, et al. (2012)	Canada	Cross-sectional	Adolescents	58	43%	ND	15.3	CAPS	X	CES-D	X
Flett, Galfi-Pechenkov, et al. (2012)	Canada	Cross-sectional	University students	246	63%	ND	20.3	HMPS	X	CES-D	X
Flett, Hewitt, et al. (2012)	Canada	Cross-sectional	High school students	250	57%	ND	15.98	CAPS	X	CES-D	X
Flett, Coulter, et al. (2012)	Canada	Cross-sectional	Primary students	88	55%	ND	12.99	CAPS	SASC-R	X	X
Flett et al. (2016)	Canada	Cross-sectional	University students	191	55%	ND	20.5	HMPS	X	CES-D	X
Fredrick et al. (2017)	USA	Cross-sectional	Adolescents	168	50%	11–15	12.59	APS-R	RCMAS	CDI	X
Frost et al. (1994)	USA	Cross-sectional	University students with or without OCD	44	100%	17–23	ND	FMPS	X	X	MOCI, CACR, OTC
Graham et al. (2010)	Canada	Longitudinal	University students	240	83%	ND	20	FMPS, HMPS	X	CES-D	X
Grenon et al. (2020)	Canada	Longitudinal	High school students	648	51%	12–14	ND	PANPS	CBC	X	X
Harris et al. (2008)	USA	Cross-sectional	University students	96	70%	ND	21.21	FMPS	X	BDI	X
Hewitt et al. (2014)	Canada	Cross-sectional	Clinically depressed adolescents	55	75%	13–19	15.53	CAPS	X	BDI	X
Hewitt et al. (2002)	Canada	Cross-sectional	Children	114	61%	10–15	12.3	CAPS	CMAS	CDI	X
Hewitt and Flett (1991)	Canada	Cross-sectional	University students	104	69%	ND	22.1	HMPS	SCL	SCL	SCL
Hewitt et al. (2022)	Canada	Cross-sectional	Children	107	62%	8–15	11.5	CAPS	X	PHQ	X
Hibbard and Davies (2011)	USA	Cross-sectional	Universitystudents	231	74%	ND	20.62	FMPS	X	BDI	X
Hinterman et al. (2012)	USA	Cross-sectional	University students	232	49%	ND	19.6	PANPS	X	CCDAS	X
Núñez-Peña and Bono (2021)	Spain	Intervention	University students	251	84%	ND	F = 21.74, M = 21.08	FMPS	SMARS	X	X
Karababa (2020)	Turkey	Cross-sectional	Adolescents	639	55%	11–14	12.8	PANPS	STAI	X	X
Kenney Benson and Pomerantz (2005)	USA	Intervention	Children	104	50%	7–10	8.21	CAPS	X	CES-D	X
J. Klibert et al. (2014)	USA	Cross-sectional	University students	413	66%	ND	20.3	HMPS	BSI	BSI	X
J. J. Klibert et al. (2005)	USA	Cross-sectional	University students	475	71%	ND	20.9	HMPS	CCAS	BDI	X
La Rocque et al. (2016)	Canada	Cross-sectional	University students	503	85%	ND	18.26	HMPS	X	BDI	X
Lee et al. (2020)	China	Cross-sectional	High school students	605	53%	16–18	ND	HMPS	X	CES-D	X
Leenaars and Lester (2007)	USA	Cross-sectional	University students	117	72%	ND	23.2	APS-R	X	CES-D	X
Levine et al. (2019)	Canada	Longitudinal	High school students	174	63%	12–17	14.54	FMPS	CCAS	BSD	X
Liu et al. (2022)	China	Cross-sectional	University students	454	36%	16–21	18.51	APS-R	X	BDI	X
Luyckx et al. (2008)	Belgium	Cross-sectional	University students	192	79%	ND	18.2	FMPS	X	CES-D	X
Macedo et al. (2017)	Portugal	Cross-sectional	University students	258	80%	ND	19.26	FMPS, HMPS	POMS	POMS	X
Mackinnon et al. (2014)	Canada	Cross-sectional	University students	165	76%	ND	20.75	HMPS	DSA	DDM	X
Mackinnon et al. (2012)	Canada	Cross-sectional	University students	452	50%	ND	21.66	HMPS, FMPS	X	CES-D	X
Magson et al. (2019)	Australia	Cross-sectional	Primary students	510	50%	ND	11.2	CAPS	SCAS	SMFQ	X
Martin et al. (1996)	Canada	Cross-sectional	University students	179	60%	ND	20.54	HMPS	X	BDI	X
McGrath et al. (2012)	Canada	Cross-sectional	University students	240	83%	ND	20	DEQ, EDI-P, FMPS	X	CES-D	X
McWhinnie et al. (2009)	Canada	Cross-sectional	Children	140	51%	6–14	ND	CDAS	X	CDI	X
Menatti et al. (2013)	USA	Cross-sectional	University students	167	100%	ND	19.16	FMPS	SIAS	X	X
Mohammadian et al. (2018)	Iran	Cross-sectional	University students	131	77%	ND	22.4	FMPS	LSAS	X	X
Moser et al. (2012)	USA	Cross-sectional	Twins	292	100%	ND	20.92	FMPS	STAI	X	X
Nakano (2009)	Japan	Cross-sectional	University students	249	100%	ND	18.4	APS-R	X	SRDS	X
Nealis et al. (2020)	Canada	Intervention	University students	127	72%	ND	21	HMPS, FMPS, DEQ	X	CES-D	X
Nepon et al. (2011)	Canada	Cross-sectional	University students	155	81%	ND	20.7	HMPS	LSAS	CES-D	X
O’Connor et al. (2010)	UK	Longitudinal	Adolescents	737	50%	ND	15.2	CAPS	HADS	HADS	X
Ortega et al. (2014)	USA	Cross-sectional	Latino students	207	76%	ND	21.13	APS-R	STAI	CES-D	X
Patock-Peckham and Corbin (2019)	USA	Cross-sectional	University students	419	79%	ND	20.19	APS-R	X	CES-D	X
Plominski and Burns (2018)	USA	Cross-sectional	University students (honours and non-honours)	1027	72%	ND	Honours = 20.7, non-honours = 21.3	PANPS	STAI	CES-D	X
Reyes et al. (2015)	USA	Cross-sectional	Adolescents	173	62%	13–17	15.09	CAPS	X	BDI	X
Rice et al. (2006)	USA	Cross-sectional	University students	403	56%	17–20	ND	APS-R	X	CES-D	X
Rice and Pence (2006)	USA	Cross-sectional	University students	308	78%	ND	20.15	APS-R, FMPS, HMPS	X	X	BSI
Rice, Ashby, et al. (2007)	USA	Cross-sectional	University students	207	74%	ND	19.38	APS-R, FMPS, HMPS	X	CES-D	X
Rice, Leever, et al. (2007)	USA	Cross-sectional	Adolescents	141	57%	12–16	13.25	AMPS	X	RADS	X
Rice et al. (2008)	USA	Cross-sectional	Chronically ill adolescents	94	63%	12–17	13.5	APS-R	X	CES-D	X
Rousseau et al. (2018)	Israel	Cross-sectional	Children	180	61%	ND	8.1	FMPS	RCMAS, SCAS	X	X
Rudolph et al. (2007)	Canada	Cross-sectional	University students	100	82%	ND	20.2	HMPS	X	CES-D	X
D. L. Sherry et al. (2015)	Canada	Cross-sectional	University students	269	48%	ND	19.26	HMPS	X	BDI	X
S. B. Sherry et al. (2013)	Canada	Longitudinal	University students	127	78%	18–24	18.31	APS-R	X	CES-D	X
S. B. Sherry et al. (2014)	Canada	Cross-sectional	University students	302	73%	ND	20.84	HMPS, FMPS, DEQ	MASQ	MASQ	X
Shu et al. (2020)	Australia	Cross-sectional	Community sample	267	100%	14–19	17.8	CPQ, FMPS	RCADS	RCADS	X
Shumaker and Rodebaugh (2009)	USA	Cross-sectional	University students	511	70%	ND	19.3	FMPS, APS	SIAS, BSAM	X	X
Sironic and Reeve (2012)	Australia	Cross-sectional	High school students	105	59%	ND	17.4	APS-R	DASS	DASS	X
Sironic and Reeve (2015)	Australia	Cross-sectional	High school students	938	62%	ND	16.26	FMPS, APS-R, CAPS	DASS	DASS	X
Smith et al. (2017)	China and Canada	Cross-sectional	University students	1034	76%	ND	19.1	HMPS, FMPS	DASS	DASS	X
Smith, Vidovic, et al. (2018)	Canada	Longitudinal	University students	143	79%	ND	19.9	HMPS	X	BDI	X
Soenens et al. (2008)	Belgium	Longitudinal	Adolescents	677	50%	15–18	15.65	FMPS	X	CES-D	X
Soreni et al. (2014)	Canada	Cross-sectional	Adolescents with OCD	94	52%	9–17	13.2	CAPS	X	CDI	Child Y-BOCS
Steffen (2014)	USA	Cross-sectional	University students	1025	63%	ND	20.4	FMPS	STAI	BDI	X
Stoeber et al. (2013)	UK	Cross-sectional	University students	210	78%	ND	20.5	APS-R	X	CES-D	X
Stornaes et al. (2019)	Norway	Cross-sectional	Adolescents	832	53%	13–14	ND	FMPS, CAPS	RCADS	RCADS	X
Sturman et al. (2009)	Canada	Cross-sectional	University students	170	100%	ND	19.9	HMPS, FMPS	X	CES-D	X
Tan and Chun (2014)	Singapore	Cross-sectional	High school students	225	100%	14–15	ND	APS-R	AEQ	X	X
Tong and Lam (2011)	China	Cross-sectional	Children	59	48%	9–11	9.98	CAPS	X	CES-D	X
Vicent, Gonzálvez, et al. (2019)	Ecuador	Cross-sectional	High school students	1588	39%	12–18	14.83	CAPS	IAEP	X	X
Vicent et al. (2020)	Ecuador	Cross-sectional	High school students	1562	54%	12–17	14.83	CAPS	DASS	DASS	X
Vicent, Inglés, et al. (2019)	Spain	Cross-sectional	Primary students	1809	50%	8–11	9.53	CAPS	IAEP	X	X
Walker et al. (2015)	USA	Cross-sectional	University students	128	100%	18–23	ND	FMPS	STAI	BDI	X
K. T. Wang et al. (2016)	Russia	Cross-sectional	University students	183	67%	19–221	20.4	APS-Short	DASS	DASS	X
K. T. Wang et al. (2019)	Russia	Cross-sectional	University students	169	70%	18–23	19.6	APS-Short	DASS	DASS	X
K. T. Wang et al. (2009)	China	Cross-sectional	High school students	509	39%	14–21	16.6	APS-R	X	CES-D	X
Y. Wang and Zhang (2017)	China	Cross-sectional	University students	236	55%	ND	20.16	APS-R, FMPS, HMPS	STAI	BDI	X
Weiner and Carton (2012)	USA	Cross-sectional	University students	172	49%	ND	20.01	FMPS	TAS	X	X
Wiebe and McCabe (2002)	Canada	Cross-sectional	University students	48	100%	18–22	20	HMPS	BAI	BDI	X
Wimberley and Stasio (2013)	USA	Cross-sectional	University students	113	58%	ND	18.1	FMPS	POMS	POMS	X
Ye et al. (2008)	USA	Cross-sectional	Children with clinical OCD diagnosis	31	42%	7-18 yrs	11.77	AMPS	X	CDI	YBOCS
Yoon and Lau (2008)	USA	Cross-sectional	University students	140	79%	ND	19.8	FMPS	X	BDI	X
You and Yoo (2021)	Korea	Cross-sectional	University students	420	64%	ND	21.63	HMPS	X	CES-D	X
Zhang and Cai (2012a)	China	Cross-sectional	University students	412	55%	17–24	20.16	APS-R, FMPS, HMPS	X	BDI	X
Zhang and Cai (2012b)	China	Cross-sectional	University students	570	52%	17–24	19.8	APS-R	X	BDI	X
Zhang and Cai (2012c)	China	Cross-sectional	University students	292	55%	17–24	20.04	APS-R, FMPS, HMPS	X	BDI	X

Note: ND: statistics not reported; X: not applicable; M: Male; F: Female; CAPS: Children and Adolescent Perfectionism Scale; HMPS: Hewitt and Flett Multidimensional Perfectionism Scale; FMPS: Frost Multidimensional Perfectionism Scale; PANPS: Positive and Negative Perfectionism Scale; APS-R: Almost Perfect Scale—Revised; AMPS: Adaptive/Maladaptive Perfectionism Scale; DEQ: Depressive Experiences Questionnaire; EDI-P: Eating Disorder Inventory—Perfectionism subscale; CDAS: Children’s Dysfunctional Attitudes Scale; LSAS: Liebowitz Social Anxiety Scale; TAS: Test Anxiety Scale; BAI: Beck Anxiety Inventory; RCMAS: Revised Children’s Manifest Anxiety Scale; GTAI: German Test Anxiety Scale; STAI: State Trait Anxiety Inventory; RAAS: Revised Adult Attachment Scale; SIAS: Social Interaction Anxiety Scale; MeMa: Evaluative metacognition, negative attitudes and mathematics anxiety; SCARED: Screen for Child Anxiety Related Emotional Disorders; SASA: Social Anxiety Scale for Adolescents; SCAS: Spence Children’s Anxiety Scale; CTS: Cognitive Test Anxiety Scale; SCAS-R: Social Anxiety Scale for Children—Revised; CBC: Child Behaviour Checklist; CMAS: Children’s Manifest Anxiety Scale; SCL: Symptom Checklist-90; SMARS: Shortened Math Anxiety Rating Scale; BSI: Brief Symptom Inventory; CCAS: Costello-Comry Anxiety Scales; POMS: Profile of Mood States; DSA: Daily Social Anxiety; HADS: Hospital Anxiety and Depression Scale; RCADS: Revised Child Anxiety and Depression Scale; BSAM: Brief State Anxiety Measure; AEQ: Academic Emotions Questionnaire; IAEP: School Anxiety Inventory; BDI: Beck Depression Inventory; CDI: Children’s Depression Inventory; DASS-21: Depression, Anxiety and Stress Scale; KDI: Kandel Davies Inventory; CES-D: Centre for Epidemiological Studies Depression Scale; DEQ: Depressive Experiences Questionnaire; SMFQ: Short Mood and Feelings Questionnaire; PHQ: Patient Health Questionnaire; BSD: Brief Screen for Depression; DDM: Daily Depressed Mood; SRDS: Self-Rating Depression Scale; RADS: Reynolds Adolescent Depression Scale; MASQ: Mood and Anxiety Symptoms Questionnaire; YBOCS: Yale Brown Obsessive Compulsive Scale; MOCI: Maudsley Obsessive Compulsive Inventory; CACR: Compulsive Activity Checklist Revised; OTC: Obsessive Thoughts Checklist.

Quality assessment

Study quality ranged from 33.33% to 92.86% (M = 61.84%, SD = 7.88%), indicating that average quality was fair (see supplementary materials for quality assessment ratings for each study). With most studies being cross-sectional in nature, quality was reduced in terms of timing of the exposure, order of measuring the exposure and outcome, number of times exposure was measured and the timeframe to see an association. Almost all studies failed to report a justification for their sample size. Most studies clearly identified their research question or objective and defined their study population and measures well.

Associations of perfectionism

Pooled correlations between symptoms of anxiety, depression, OCD, and perfectionism can be seen in Table 2. There were some outliers where studies had smaller pooled correlations, removing these outliers made no difference in the results, so they were retained, see Table 2.

Table 2. Pooled correlations between perfectionism dimensions and symptoms.

	Perfectionistic concerns	Perfectionistic strivings
Anxiety
Pooled r [95% CI]	0.373 [0.307,0.439]**	0.054 [0.010,0.098]*
Heterogeneity [I^2 %]	Q(52) = 2431.086 [97.735%]**	Q(30) = 184.922 [82.87%]*
Egger’s z	−2.715*	−1.242
Pooled r without outlier [95% CI]	0.389 [0.326,0.452]**	0.066 [0.027,0.104]**
Test Anxiety
Pooled r [95% CI]	0.405 [0.306,0.504]**	−0.058 [−0.330,0.214]
Heterogeneity [I^2 %]	Q(7) = 107.155 [91.96%]**	Q(5) = 422.711 [98.268%]**
Egger’s z	0.121	0.617
Pooled r without outlier [95% CI]	0.440 [0.368,0.511]**	0.067 [−0.017,0.150]
Depression
Pooled r [95% CI]	0.400 [0.349,0.450]*	0.006 [−0.049,0.061]
Heterogeneity [I^2 %]	Q(94) = 3897.930 [97.360%]**	Q(61) = 1097.842 [94.285%]**
Egger’s z	−2.832*	0.142
Pooled r without outlier [95% CI]	No outliers	0.018 [−0.022,0.058]
OCD
Pooled r [95% CI]	0.420 [0.362,0.478]**	0.192 [0.109,0.275]**
Heterogeneity [I^2 %]	Q(5) = 4.334 [0.00%]	Q(3) = 4.905 [16.802%]
Egger’s z	0.529	1.744
Pooled r without outlier [95% CI]	No outliers	0.224 [0.137,0.312]**

Note: *.05, **.001; CI = confidence interval.

Anxiety symptoms

There was a significant medium pooled correlation between perfectionistic concerns and symptoms of anxiety (r = .37, 95% CI [.31, .44], p < .001). Due to high heterogeneity (I²  = 97.74%; Q (52) = 2431.09, p < .001), a moderator analysis was conducted, and although age was a significant moderator (β = .18, 95% CI [.02, .35], p = .032), heterogeneity was still high; I²  = 97.47%, Table 3). Specifically, the correlation was weaker for children (r = .22, 95% CI [.03, .42], p = .023) compared to adolescents and young adults (r = .41, 95% CI [.34, .47], p < .001). Further, country income status was not a significant moderator (β = −.07, 95% CI [−.20, .07], p = .310). There was significant publication bias (z = −2.715, p = .007).

Table 3. Moderator analysis results for age and low-middle income country.

	B [95% CI]	SE
Perfectionistic concerns
Anxiety—Age moderator
Intercept	0.226[0.077, 0.374]*	0.076
Age (O13)	0.180[0.016, 0.345]*	0.084
Heterogeneity	Q(51) = 2225.135, p < .001, I^2 = 97.474%
Anxiety—Country Income status
Intercept	0.424 [0.310, 0.538]***	0.058
Income (High)	−0.070[−0.204, 0.065]	0.069
Heterogeneity	Q(49) = 1847.434, p < .001I^2 = 96.981%
Test Anxiety—Country income status
Intercept	0.433 [0.277,0.590]**	0.080
Income (High)	−0.047[−0.250,0.155]	0.103
Heterogeneity	Q(6) = 106.144, p < .001, I^2 = 91.336%
Depression—Age moderator
Intercept	0.322 [0.171, 0.472]**	0.077
Age (O13)	0.088 [−0.072, 0.247]	0.081
Heterogeneity	Q(93) = 3762.230, p < .001, I^2 = 97.272%
Depression—Country income status
Intercept	0.441 [0.345, 0.537]**	0.049
Income (High)	−0.047 [−0.154, 0.061]	0.055
Heterogeneity	Q(90) = 2676.277, p < .001, I^2 = 95.918%
Perfectionistic strivings
Anxiety
Intercept	0.055[−0.025, 0.135]	0.041
Income (High)	−0.008[−0.105, 0.089]	0.049
Heterogeneity	Q(28) = 171.549, p < .001 , I^2 = 81.524%
Test anxiety
Intercept	−0.095 [−0.474, 0.284]	0.193
Income (High)	0.075 [−0.466, 0.615]	0.276
Heterogeneity	Q(4) = 416.908, p < .001, I^2 = 98.350%
Depression
Intercept	0.009 [−0.105,0.123]	0.058
Income (High)	0.000 [−0.131,0.131]	0.067
Heterogeneity	Q(58) = 1055.371, p < .001, I^2 = 94.143%

Note: *.05, **.001; CI = confidence interval; Age (O13) = age coded for over 13 and under 13 samples (comparison); SE = Standard error.

There was a significant very small pooled correlation between perfectionistic strivings and anxiety symptoms (r = .05, 95% CI [.01, .10], p = .015), with no significant publication bias (z = −1.242, p = .214). There was high heterogeneity (I²  = 82.87%; Q (30) = 184.92, p < .001), however, in a meta-regression country income status was not a significant predictor (β = −.01, 95% CI [−.105, .089], p = .872; see Table 3, Figures 2 and 3).

Figure 2. Combined forest plot of the pooled correlations with perfectionistic concerns.

Note: Centre dot denotes the pooled correlation (r).

Figure 3. Combined forest plot of the pooled correlations with perfectionistic strivings.

Note: Centre dot denotes the pooled correlations.

Test anxiety symptoms

There was a significant medium pooled correlation between perfectionistic concerns and symptoms of test anxiety (r = .41, 95% CI [.31, .50], p < .001). Publication bias was not significant (z = 0.121, p = .903). Heterogeneity was high (I²  = 91.96%; Q (7) = 107.16, p < .001), and country income status was found not to be a significant moderator (β = −.05, 95% CI [−.250, .155], p = .646).

There was no significant association between perfectionistic strivings and symptoms of test anxiety (r = −.06, 95% CI [−.33, .21], p = .674). Publication bias was not significant (z = 0.617, p = .538). A meta-regression was conducted due to high heterogeneity (I²  = 98.27%, 95%; Q (30) = 184.92, p < .001), and country income status was not a significant moderator (β = .075, 95% CI [−.466, .615], p = .786).

Depression symptoms

There was a significant medium pooled correlation between perfectionistic concerns and depressive symptoms (r = .40, 95% CI [.35, .45], p < .001). There was significant publication bias (z = −2.832, p = .005). Due to high heterogeneity (I²  = 97.36%; Q (94) = 3897.93, p < .001), a meta-regression was performed, and age was not a significant moderator (β = .09, 95% CI [−.07, .25], p = .28), nor was country income status (β = −.05, 95% CI [−.15, .06], p = .40)

There was no significant association between perfectionistic strivings and symptoms of depression (r = .01, 95% CI [−.05, .06], p = .830). Publication bias was not significant (z = 0.142, p = .887). Heterogeneity was high (I²  = 94.29%; Q (61) = 1097.84, p < .001); however, country income status was not a significant predictor (β = .00, 95% CI [−0.13, 0.13], p = .995) (see Table 3).

OCD symptoms

There was a significant medium pooled correlation between perfectionistic concerns and OCD symptoms (r = .42, 95% CI [.36, .48], p < .001). Publication bias was not significant (z = 0.529, p = .596). There was low heterogeneity (I²  = 0.00%; Q (5) = 4.33, p = .502). There was a significant small correlation between perfectionistic strivings and OCD symptoms (r = .19, 95% CI [.11, .28], p < .001). There was low heterogeneity (I²  = 16.80%; Q (3) = 4.91, p = .179) and publication bias was not significant (z = 1.744, p = .081). Considering the low levels of heterogeneity and the small number of studies for the associations between perfectionism and OCD symptoms, we did not run any moderator analyses (Baker et al., 2009).

Discussion

This was the first meta-analysis to focus on the association between perfectionism and symptoms of anxiety, OCD, and depression specifically among young people. Perfectionistic concerns had medium, pooled correlations with symptoms of anxiety, OCD, test anxiety and depression among young people. Perfectionistic strivings had smaller, though significant associations with symptoms of anxiety and OCD in young people. There was a weaker relationship between perfectionistic concerns and anxiety in children compared to adolescents and young adults, however age was not a significant moderator of the other relationships investigated.

The findings support previous meta-analyses in adult populations, with similar medium pooled correlations between perfectionistic concerns and symptoms of anxiety (current r = .37 vs r = .35), OCD (current r = .42 vs r = .30; Limburg et al., 2017) and depression (current r = .40 vs r = .39). Hence, our findings confirm in youth the consistent association between perfectionistic concerns, and symptoms of anxiety, OCD, and depression. Although the data were cross-sectional, so cannot provide support for causality between perfectionism and psychopathology, the results are in line with the cognitive-behavioural model of perfectionism, where self-worth based on striving is associated with adverse consequences, including symptoms of anxiety and depression (Shafran et al., 2002). Shafran et al. (2023) recently summarised evidence in support of the model, for example, experimental studies demonstrating the causal link between the induction of perfectionism leading to an increase in negative affect (e.g. Hummel et al., 2023; Yiend et al., 2011). Further research on the efficacy of CBT-P in youth is warranted to add to existing literature showing the treatment reduces symptoms and, prevents onset of, anxiety and depression in adolescents (e.g. Nehmy & Wade, 2015; Shu et al., 2020). This is pertinent considering while self-oriented and socially prescribed perfectionism are significantly reduced following standard CBT for anxiety in clinically anxious children, self-oriented perfectionism was a predictor of poorer treatment outcome at post-treatment and 6-month follow-up (Mitchell et al., 2013).

The pattern of results for perfectionistic strivings was interesting. While our results were similar to Limburg et al. (2017) for strivings and OCD symptoms (current r = .19 vs r = .14), we found no relationship between strivings and depressive symptoms (current r = .01 vs r = .11). We also found a smaller, though significant, association between strivings and symptoms of anxiety (current r = .05 vs r = .14) compared to Limburg et al. (2017), but similar to Smith, Sherry, et al. (2018)’s significant, and also small, association (r = .08). A potential reason for the smaller association with symptoms and perfectionistic strivings in the current review is the sample populations. In our review 3% of the studies were clinical samples versus 35% in Limburg et al. (2017). Potentially, in clinical samples perfectionistic strivings may be more problematic than in non-clinical samples. Our findings do not support the notion that perfectionistic strivings are positive (Stoeber & Otto, 2006). Our findings in young people echo those of reviews in adults (Limburg et al., 2017), perfectionistic strivings are associated with symptoms of psychopathology. Nevertheless, the association between strivings and psychopathology was smaller than for concerns, and we did not find a relationship between perfectionistic strivings and symptoms of test anxiety or depression.

The results from the moderation analyses indicated the relationship between perfectionistic concerns and symptoms of anxiety was weaker in children compared to adolescents and young adults. However, given age was not a significant moderator of any of the other relationships investigated, it is unclear whether age effects in children compared to adolescents and young adults are pertinent in the relationship between perfectionism and psychological symptoms. Further meta-analyses investigating age as a moderator between perfectionism dimensions and psychological symptoms are required to determine if this result is replicated and to further investigate any potential reasons for this difference.

There were several limitations. First, most samples were non-clinical, meaning the results cannot be generalised to clinical populations. We were unable to examine clinical versus non-clinical groups as a moderator due to insufficient studies, and this should be an aim in a future meta-analysis. Future research should address this gap by conducting further research in the area of perfectionism in young people in clinical populations. Second, there were only six studies examining the relationship between perfectionism and symptoms of OCD in young people, hence the conclusions should be interpreted tentatively. Third, in the perfectionism literature, there are a lack of longitudinal studies. Given this review was cross-sectional, we cannot infer causality in the association between perfectionism and symptoms of anxiety, depression, and OCD in young people. Future research should employ longitudinal designs, specifically following youth from adolescents into adulthood and examining impact of perfectionism in development of symptoms of anxiety, depression, and OCD. Finally, we were unable to examine gender due to insufficient data, future meta-analyses should aim to do this to inform any potential gender differences in the relationship between perfectionism and psychopathology in young people.

In conclusion, perfectionistic concerns had a positive association with psychopathology in youth, and perfectionistic strivings had a smaller, though significant relationship with some aspects of anxiety and OCD symptoms. Most treatment research has focused on adults, research should now focus on treatment of perfectionism in youth to provide effective intervention as early as possible to prevent and treat symptoms of anxiety, depression, and OCD.

Disclosure statement

Sarah Egan receives royalties for the books ‘Cognitive-Behavioural Treatment of Perfectionism’ and ‘Overcoming Perfectionism: A Self-Help Guide Using Scientically Supported Cognitive Behavioural Techniques’.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/16506073.2023.2211736

Correction Statement

This article has been corrected with minor changes. These changes do not impact the academic content of the article.

Notes

1. One study recruited participants form both China and Canada therefore is included in the count of both high- and low-income studies but not included in the moderator analyses.