Underestimation of overweight weight status in children and adolescents aged 0‐19 years: A systematic review and meta‐analysis

Abstract Background Perceptions of children's weight status may be important in obesity prevention and treatment. Aims This review identifies the prevalence of the underestimation of overweight status in children by parents/main carers, children, and healthcare professionals (HCP). The review critically synthesized both quantitative and qualitative evidence to explore the factors associated with this underestimation. The diverse methods used to assess this phenomenon are reported. Methods Pooled effect sizes were calculated using random‐effects model. Published studies, up to 2020, were accessed using the following search engines: CINAHL, EMBASE, PUBMED, and Psych‐Info and including the “Cited by” and “Related Articles” functions. Hand‐searching was used to retrieve further articles. Publication language and location had no bearing on the nature of the included studies. Results A total of 91 articles were included. In the quantitative studies, 55% (95% CI 49%–61%) of caregivers underestimated their child's level of overweight and obesity using a verbal scale and 47% (95% CI 36%–55%) using visual scales. Of the children studied, 34% (95% CI 25%–43%) underestimated their own level of overweight and obesity using both scales. In (n = 3) articles, HCPs reflected this misperception, but limited studies prevented meta‐analysis. Underestimation was associated with the child's age, gender, BMI and parental weight status, ethnicity and education. In the qualitative studies, parents/main carers of children with overweight and obesity described their child's weight in terms other than overweight, for example, “big boned,” “thick,” and “solid.” Conclusion The results confirm the prevalence of underestimation of child overweight status across international studies. Understanding the factors which lead to this inaccuracy may help to improve communication within the therapeutic triad and facilitate the recognition and management of children's overweight status.


| INTRODUCTION
Levels of childhood overweight and obesity continue to increase globally. 1 Approximately 41 million children aged 0-5 years are classified as overweight or obese, while 340 million children aged 5-19 years can be categorized as overweight or obese. 2 Childhood overweight is a highly complex condition with biological, behavioral, social, economic, environmental, and cultural causes. 3 The widespread nature of childhood overweight is a cause for concern since it is associated with a range of negative medical and psychosocial effects, both short and long-term. 4 Changes in diet, physical activity levels and, more recently sedentary activity are internationally considered to underpin interventions for the management of both adult and childhood obesity. 5 To enact behavior change it is necessary to achieve concordance which, in the case of childhood overweight and obesity, involves the therapeutic triad of the child (if sufficient cognitive capacity), their parents/main carers, and healthcare professionals (HCP) involved in their care. Effective weight control interventions for children are strongly informed by the degree to which the overweight and obesity perceptions of children, parents/main carers and HCP are accurate and consistent with one another. 6 It is, therefore, essential that weight-related interventions are applied using an understanding of and effective working with key stakeholders and their perceptions of child overweight status. 7 It is generally accepted that a significant barrier to intervention is that parents underestimate their children's overweight status. 8 Furthermore, there is some limited evidence that children themselves and HCP may also underestimate. 9 In-depth engagement and cooperation across the therapeutic triad is required for concordance-a dynamic which would be threatened by misaligned perceptions of the situation. 6 The psychological resistance that sometimes arises in parents due to the perceived link between childhood obesity, stigma, and blame, may result in their denial of a HCP's diagnosis, which decreases their receptiveness to the proposed interventions. 10 Data from several studies suggest that negative emotions like anger and shame may lead to resistance in the form of refusing to take advice or ignoring referral appointments. Previous research has demonstrated that HCP can perceive conversations around weight to be "difficult" and thus avoid raising the subject at all. 11 Understanding how sociodemographic, anthropometric, and cognitive factors, that may influence underestimation, may help to preempt these "difficult" conversations, facilitate the recognition of children's overweight status, unify the therapeutic triad and help to notify the improvement of future weight-related communications.
Earlier reviews in this area focused on parental perceptions toward child weight and repeatedly demonstrated that parents tend to underestimate their child weight. [12][13][14] The most recent meta-analysis searched databases up to 2013, reporting that 50.7% of parents underestimate their children's weight if they are living with overweight or obesity. 5 The current systematic review and meta-analysis aimed to update our knowledge by statistically identifying the prevalence of the underestimation of children's overweight status in children aged ≤19 years old in a wider sample (parents/main carers, children themselves and HCP). In addition, this review critically synthesized both quantitative and qualitative evidence to extend our understanding by investigating the factors associated with this underestimation. The systematic review also examined the diversity of methods used to assess this phenomenon. overweight, obesity, weight, and underestimations." Boolean operators were used to group the terms (see Table 1), thereby facilitating a more wide-ranging search. In addition, a reference list of the identified literature was then hand-searched and the "Cited by" and "Related Articles" functions in the search databases were also used to retrieve further relevant articles. Publication language and location had no bearing on the nature of the included studies. The participants were the parents/main carers of children with overweight/obesity, children with overweight/obesity or HCP. It should be noted that data were only on parents/main carers' assessments of their own children and not assessments of other children. Similarly, the children's assessment were of themselves and not other children, whereas the HCP 0 assessment was generalized. Exclusion criteria included studies that do not provide anthropometric measurements with which to compare perceptions, as well as studies focused on perceptions toward "underweight" or "normal weight" participants only.

| Systematic review registration
The protocol has been registered at the International Prospective Register of Systematic Reviews (PROSPERO); registration number:

| Quality assessment
A modified Newcastle-Ottawa Scale adapted for single use in a specific context 15 was used to assess the quality of the 91 included studies. Following the authors' guidelines each study was assessed on the following criteria: methods used to select participants (i.e., sampling bias), methods used for controlling confounding (i.e., performance bias), statistical methodology (i.e., detection bias), and what the methods used to measure the outcome variables are (i.e., information bias). This scale measures the risk on a scale of 0 (high risk of bias) to 3 (low risk of bias) (see Table 2).

| Statistical methods
After the extraction of quantitative data, the main effect sizes for the percentage of underestimation by parents/main carers and children using both visual and verbal scales were determined using STATA 14 and a meta-analysis was performed using R program (There was insufficient data to consider underestimation in HCP). For each study, the effect size was calculated for the reported adjusted proportion of the participants who underestimated children's overweight weight status. A random-effects meta-analysis model was used in the study since the studies stem from different populations and design-related heterogeneity. 105 To test heterogeneity between-studies, the χ 2 with a statistic Q was used to examine the null hypothesis that all studies have the same true effect, 36 namely H0: τ 2 = 0. Given that the Q-test sometimes may fail to detect heterogeneity when the number of samples is small (i.e., less than 10), 106 the statistics I 2 was also used to assess the proportion of the observed heterogeneity, 107 as it is not affected by sample size. I 2 values range from 0% to 100% with three levels of I 2 : low, moderate, and high corresponding to values of 25%, 50%, and 75%. 106 In addition, the 95% CI around the I 2 statistic was also calculated to determine whether the heterogeneity is present.
Overall, τ 2 , Q-test and I 2 were used in this research to decide whether the estimated effects are homogeneous. Thereafter, pooled effect sizes were estimated via random effects models, since the studies are based on different populations. To visualize the overall estimates of the study effects with corresponding confidence intervals, forest plots were used. 108 Sensitivity analyses of the meta-analyses were conducted using a range of visual approaches including difference in values (DFFITS), Cook's distances, studentized deleted residuals, Cov-ratios, and estimates of heterogeneity. This identified some potentially outlying studies, but with a low impact on overall estimate sizes, so they were retained in the meta-analyses. In evaluating heterogeneity among the studies, it was apparent that there was a significant influence in all data-sets, therefore, random effects models were applied to all studies. Q2 "In comparison with other children his/her age' with regard to weight" A1 "dis-agree a lot," "disagree a little," "no strong feelings either way," "agree a little," and agree a lot" A2 "much thinner" to "much more overweight" A community-based cohort study recruited from maternal and child health centres Q1 "I am worried my child is overweight right now" Q2 "In comparison with other children his/her age' with regard to weight" A1 "dis-agree a lot," "disagree a little," "no strong feelings either way," "agree a little," and "agree a lot." A2 "much thinner" to "much more overweight" Consecutive sampling from paediatric practices Q1 "I feel my child is" Q2 "did your child's doctor ever tell you that your child is gaining weight too fast or is overweight?"

Q3
We also sked the participants to mark the drawing that looked the most like their child A1 "underweight," "a little underweight," "about the right weight," "a little overweight," or "overweight" A2 "yes," "no," or "I don't know" A3 Collins' Figure  A1 "much higher than normal," "higher than normal," "normal," "lower than normal" or "much lower than normal."

participants Mothers
Connivance sample was from waiting room of health care facility Q1 "I feel my child is" Q2 "has your child ever been told by a doctor that he/ she is at-risk for overweight, overweight, or obese?" A1 "underweight," "normal weight," "a little overweight," or "overweight"

| Objective measures
Across the studies, all were based on weight to height ratios with different criteria for categorization to determine a child's weight -787

| Subjective measures
Across 80 of the studies, the participants verbally classified their children's or their own weight status with a forced-choice rating scale, whereby participants described their weight status by selecting a category that was based on medicalized weight related standards.
In 20 cases, a numerical Likert-type scale was employed, in which a child's weight was considered along a spectrum, from very underweight to very overweight. Across the included studies, differing terms were used to describe overweight and obesity (e.g., "underweight" to "obese," "underweight" to "overweight," "somewhat thin" to "somewhat heavy," "markedly underweight" to "markedly overweight" and "thin" to "fat"). One study relied on a visual analogue scale, where caregivers were requested to mark the spot on the line correlated with their child's weight status. 38 In 13 studies, visual evaluations of sketches, silhouettes, photo- reported that mothers tended to have more precise views of their children's nutritional status when selecting silhouettes corresponding to BMI, rather than when using multiple-choice questions. Although Moschonis et al. 72 reported that mothers' assessments of their children's weight status were more accurate when using verbal rather than visual classification tools, the study relied on outdated NCHS growth charts from 1979 which depicted only pre-pubescent children. Chaimovitz et al. 9 also employed verbal and visual scales for the assessment of perceived weight status, the intention being to safeguard against the influence of weight-related terminologies used in the measure. The rate of weight status underestimation was comparable across the two scales, which is consistent with our metaanalysis which provides very similar estimates of the extent of underestimation by parents/carers, regardless of the use of verbal or visual scales. The same process of underestimation of overweight and obesity was also found when HCP were asked whether each child was classified as underweight, having a healthy weight, overweight, or obese. Although there was insufficient data to perform a metaanalysis, Spurrier et al. 86 Chaimovitz et al. 9 and Tarasenko, Rossen and Schoendorf 6 all reported underestimation of overweight by a significant number of HCP. These studies included general practitioners (GPs) and pediatricians as well as physicians from hospitals, medical centers, and clinics. Among the reasons proposed for why HCP may underestimate children's overweight status is that, HCP do not take weight measurements or provide weight management guidance because their patients do not seem to be overweight or obese. 111 Phenomenological methods were employed by Redsell et al. 72 Eli et al. 100 and Thompson et al. 87 in order to gain insight into parents' perceptions of their children's weight status. These studies adopted qualitative research designs with semi-structured interviews and focus groups. The results from the studies indicated that mothers conceptualized a child's weight differently and no participant held to the medical definition of childhood obesity which the researchers proposed. Instead a preschool child was considered to have a healthy weight if they could take part in a specific activity, appeared healthy, were happy, and were not bullied. 72,87,100 Eli et al. 100 reported that no participant described the weight status of preschool children as "obese" or "overweight" but used terms such as "chunky," "stout," "chubby," "stocky," "big boned," "robust," "solid," and "pudgy."

| Methodological quality
Objective outcome measures were employed in almost every study (n = 89) except one where weight and height were reported by parents. 61 In (n = 63) studies, it was not clear whether the experimental equipment was standardized and information was rarely given as to whether the objective child weight assessment process involved identical equipment across the entire sample. In the majority of studies (n = 80), data were not gathered as to whether participants already knew about the child's weight status. Weight classification methods, confounding variables, statistical analysis methods and the possibility of non-respondent bias were identified in almost all studies as issues that could have reduced the methodological quality of the studies (see Table 2). Re; the diversity of the methods used to assess underestimation of overweight status, 75% of the included studies offer no explanation or justification for using specific terminology in question and/or responses in the verbal assessment method. The methodological quality ranged from high to low quality; however, the majority were relatively high in most of the studies (see Table 2). Therefore, the results seem to give a reliable estimate of the available data on this subject.

| Meta-analyses of underestimation
The main finding produced from the analysis of the quantitative studies is that the prevalence of underestimation of overweight is high among parents/main carers and children. When parents/main carers used a verbal scale (79 studies), 55% underestimated overweight and obesity (95% CI 48%-61%) (see Figure 2). Similarly, when using a visual scale (see Figure 3), the prevalence of underestimation 788by parents was 47% (95% CI 36%-55%). Children also tended to underestimate their weight (see Figure 4) with a prevalence estimate of 34% (95% CI 25%-43%). Although there was insufficient data to perform a meta-analysis, Spurrier et al. 86 Chaimovitz et al. 9 and Tarasenko, Rossen and Schoendorf 6 reported that underestimation of overweight also applied to a significant number of HCP with a range of (20%-60%).
The studies examined a number of factors that might contribute to inaccuracy in identifying children who were affected by overweight or obesity. These factors included the child's age where the overweight status of younger children was more likely to be underestimated than older children. 32 -789 were more likely to underestimate their child's weight than relatively well educated parents. 33,41,62,83,91,98 Homogeneity earlier reviews conducted by Parry et al. 12 Doolen et al. 13 Rietmeijer-Mentink et al. 14 and Lundahl et al. 5 However, it extends these works with a meta-analysis that demonstrates that the majority of parents/ main carers (55%, 95%CI 49%-61%) underestimated their children's level of overweight, while a substantial proportion of children underestimated their overweight status (34%, 95% CI 18%-54%). HCP shared this misperception despite the low number of studies preventing a meta-analysis. This finding is in accordance with earlier literature, which shows that the underestimation of overweight is also very common in adults. 9 The findings of this review are of concern if one accepts that alignment is required within the therapeutic triad for a trusting relationship and concordance. 9  can lead to negative emotions. [8][9][10]13,29 In the qualitative studies, parents commonly describe their children in terms other than obese, such as "big boned," "thick," and "solid," and demonstrate a strong desire to avoid labelling their child with medical terminology. 100 Perceived blame, internalized stigma, and negativity may generate resistance to the language of the discussion, prevent accurate overweight status classification, and ultimately trigger reduced intention to engage in behavior change. 9,10 The effect of this on our results is that the studies, which utilized terminology associated with negative emotion, might under-report the prevalence of overweight status and yet the majority of included studies (n = 66) neglected to describe the questionnaire development process and failed to justify the weight-related terminologies used, even when in-paper references to previous uses of the questionnaire were consulted (n = 12).
Visual scales avoid the need for labeled response options and the issues involved in selecting neutral terminology. 9 However, the 13 studies using these methods are beset by the same variation in instruments (silhouettes, etc.) as the methodology utilizing verbal scales. Studies which allow direct comparison suggest that visual scales are more accurate than verbal scales. 59 In view of this consideration, studies using pictoral assessment methods for the views of parents and children result in a slightly less underestimation of overweight/obesity (47%) compared to verbal scale (55%).
This result may further suggest that parents and children/adolescents do recognize the weight status but do not verbally label it as overweight. 14 This is perhaps because of the negative association with the word overweight and the stigma attached to people with obesity in society. 8,68 However, there is little evidence using pictorial assessment method in the current review to make informed conclusions on it.
Despite the diversity in the study design discussed above, it is noteworthy that such a large number of studies from across the world, have investigated this phenomenon and the meta-analyses demon- However, Linchey et al. 62 found no association between parent weight status and underestimation. Despite the suggestion that lower levels of maternal education are associated with the underestimation of overweight in children 41,44,48,62,83,91,98 there are some inconsistencies in the evidence by Adams et al. 17 Carnell et al. 29 and He and Evans 47 suggesting no link between maternal educational levels. In contrast, Ruiter et al. 82 found that the parents who had a high level of education were more likely to underestimate their child's overweight status. Ethnicity was also positively associated with the overall underestimation of child's weight status in two studies. 63,89,91 Molina et al. 63 found white children were less likely to be underestimated than non-white. The greater acceptance of larger body size in some cultures may account for these differences. 111 In addition, parental underestimation of child's overweight and obesity has been reported to be particularly common among Hispanic individuals, 89 an ethnic group with very high level of overweight and obesity prevalence and a likelihood to identify larger body sizes as being more normal. 111 In the recent study by Linchey et al. 62  This systematic review and meta-analysis is associated with a number of strengths and limitations. A key strength is that it includes both quantitative and qualitative studies, which allows greater understanding. Over and above the existing knowledge in this area, the study sought to examine the perceptions of parents/carers, children, and HCP. This review includes participants from a range of socioeconomic and cultural backgrounds, with a wide age range of children (up to 19 years old) in both experimental and population settings.
However, an important limitation of this systematic review has been the lack of a statistical examination of HCP' perceptions, which stems from the scarcity of relevant studies in the literature. Additionally, gender-based disparities in the parental tendency to underestimate child overweight were not tested, since the majority of the included studies examined mother-to-child rather than father-to-child dyads.
Lastly, the results could suffer from publication bias due to the fact that some of the unpublished studies were not included.
This systematic review's findings indicate that when the prevalence of underestimation of child weight status is acknowledged, viable interventions can be promoted which involve parents, children, and HCP. For the purpose of raising awareness among parents of children with overweight and obesity, it is necessary to ensure that HCP themselves have an adequate understanding of these conditions. This could be facilitated if HCP engage directly with the various ways in which children and their parents define overweight, and to gain insight into the long-term consequential health factors of being categorized as overweight and obese. 116 As emphasized by Puhl et al. 10 the psychological resistance that sometimes arises in parents due to the perceived link between childhood obesity, stigma, and blame, may result in their denial of a healthcare professional's diagnosis, which decreases their receptiveness to proposed interventions. Hence, HCP should engage directly with concerns over stigma when discussing childhood overweight and obesity with parents, assuring them that they are not to blame for the condition, and that their participation in the cooperative management of their children's overweight condition represents a beneficial step forward. Additionally, HCP should refrain from engaging parents in such a way as to instill guilt or a sense of being judged. Lastly, sensitive framing of each discussion regarding child weight status should be a top priority.

| CONCLUSION
The results of this wide-reaching systematic review confirm that the underestimation of child overweight status is highly prevalent. Those that design child weight management interventions should expect that for many parents/main carers and children, identifying and problematizing their child's overweight status will be unexpected and requires extra consideration to align beliefs. It may also lead to negative emotions and resistance. Although understanding the 792 -ALSHAHRANI ET AL.
factors which lead to inaccuracy in assessing child's overweight status may help to pre-empt "difficult" conversations and facilitate the recognition of children's overweight status, much work is required.
Future studies could usefully consider a wider range of parents/main carers, child and HCP characteristics, utilize multivariate study designs, and use more consistent data collection methods with an emphasis on visual scaling.

ACKNOWLEDGMENT
Abrar Alshahrani is currently undertaking a PhD funded by the Saudi Arabian Government and is on secondment from the University of Jazan.

CONFLICTS OF INTEREST
The authors declare no conflicts of interest.

AUTHOR CONTRIBUTIONS
Abrar Alshahrani carried out the literature search, analyzed the data and wrote the manuscript. Farag Shuweihdi revised meta-analysis.
Judy Swift provided critical feedback to the manuscript. Amanda Avery provided critical feedback and helped shape the research, analysis and manuscript.