Accuracy of anthropometric indicators of obesity to identify high blood pressure in adolescents—systematic review

Background Anthropometric indicators of obesity have been associated with blood pressure in adolescents. However, the accuracy of anthropometric indicators of obesity for screening for high blood pressure (HBP) in adolescents is not known. Thus, the aim of the present study was to summarize the set of evidence regarding the predictive ability of anthropometric indicators of obesity to identify HBP in adolescents. Methods Searches were performed in five databases: MEDLINE, Web of Knowledge, Scopus, Scientific Electronic online (SciELO) and SportDiscus. The inclusion criteria for studies were: adolescents aged 10–19 years or mean age included in this range, observational and intervention studies, studies that proposed cutoff points for anthropometric indicators of obesity, and studies in English, Portuguese and Spanish. The methodological quality of studies was assessed using the QUADAS-2 instrument. Results Ten studies met the inclusion criteria and had their information summarized. Based on the information described in these studies, the anthropometric indicators body mass index (BMI), waist circumference (WC), waist-to-height-ratio (WHtR), triceps skinfold thickness, body adiposity index, C index, body mass, waist-to-arm span ratio, arm fat area, average arm perimeter, fat percentage and arm span were likely to be used in high blood pressure (HBP) screening among adolescents. However, only one study showed acceptable values (moderate to high precision) in relation to the accuracy measurements of described cutoffs. Conclusion Caution is suggested in the use of anthropometric indicators of obesity for HBP screening in adolescents, in which a greater number of studies with accurate diagnostic tools are necessary.


INTRODUCTION
Anthropometric indicators such as body mass (BM), height, waist circumference (WC) and hip circumference (HC) have been described as useful tools to detect factors associated with cardiovascular risk (Cassiano et al., 2019), such as insulin resistance, metabolic syndrome, and dyslipidemia (De Faria et al., 2009;Mastroeni et al., 2019;Beck, Da Silva Lopes & Pitanga, 2011). Additionally, the predictive power of anthropometric indicators for screening for high blood pressure (HBP) in children and adolescents (De Quadros et al., 2019;Rimárová et al., 2018) has been described, and it has been reported that anthropometric indicators of obesity such as body mass index (BMI), waist-toheight ratio (WHtR) and WC have acceptable discriminatory power to identify HBP in adolescents (Araújo, Ramos & Barros, 2019;Liew et al., 2019;De Araújo Pinto et al., 2017).
Although the measurement of blood pressure levels is an important component of health assessment routines in the pediatric population, difficulties related to the measurement (e.g., need to use specific instrument or choose the appropriate cuff for the child/adolescent's arm) or classification of measured information (e.g., insertion of pressure values in panels of growth curves) are configured as barriers for the measurement of blood pressure levels in environments with structure different from that observed in clinical centers, such as schools or sports clubs (Barroso et al., 2021). Thus, the use of anthropometric indicators can be a useful tool in the screening for HBP in children and adolescents.
In this sense, the aim of this review was to investigate the accuracy of anthropometric indicators of obesity for screening for HBP in adolescents.

MATERIAL & METHODS
The method used in this systematic review was consistent with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) (Liberati et al., 2009) statement (Supplemental 1). The review was registered in the ''International Prospective Register of Systematic Reviews'' (PROSPERO) under number CRD42020151554, and is available in full at: http://www.crd.york.ac.uk/prospero/.
remaining studies were read in full and were selected based on inclusion criteria. In case of doubts among researchers about the inclusion of articles, a senior researcher was consulted (DASS). After the selection of articles included in the review, reference lists were read with a view of identifying possible studies not identified in the systematic search.
The survey results in each database were exported to the EndNote R version X4 reference manager (Thomson ISI ResearchSoft, Clarivate Analytics, Philadelphia, PA, USA, 2010).

Search strategies, descriptors and keywords
The selection of descriptors occurred by consulting the Health Sciences Descriptors (DeCS) (Pellizzon, 2004) and Medical Subject Headings (MeSH) platforms (Dhammi & Kumar, 2014). In addition, terms, and keywords used in literature reviews and original articles were used to determine these descriptors. Thus, searches for available information were conducted considering studies in English, Spanish and Portuguese (Supplemental 2). Boolean operators ''AND'' and ''OR'' were used to relate information between groups and according to each block of information, respectively. The groups of descriptors used in the systematic search for information were the following: (1)

Data extraction and Quality assessment
The information extracted from each study was as follows: author and year of publication, study location, sample size, age group, study design, anthropometric measures and indicators, instrument used/means of measuring blood pressure (BP), BP measurement recommendations, cutoff points estimated by studies, classification adopted for the identified BP values, indicators and diagnostic information for the predicted cutoff points (area under the curve (AUC)), sensitivity, specificity, positive predictive values (PPV), negative predictive values (NPV), optimal point criteria, measures of association for predicted results. Additionally, adjusted results were extracted from included studies and, when available, were presented according to sex and age group.
Although the use of cutoff points with AUC values lower than 0.70 (Akobeng, 2007;Fischer, Bachman & Jaeschke, 2003) is not recommended, studies included in this review that reported results that anthropometric obesity indicators had good predictive capacity for HBP used cutoff points with AUC < 0.70 to support such information. For this reason, in the present study, AUCs > 0.50 to < 0.70 were considered to have low diagnostic accuracy, AUCs ≥ 0.70 to < 0.90 were considered to have moderate predictive capacity, and AUCs ≥ 0.90 were considered to have high predictive capacity for the analyzed outcome (Akobeng, 2007;Fischer, Bachman & Jaeschke, 2003).
The assessment of the methodological quality/risk of bias of studies was carried out independently by two reviewers (LLB & TRL) using the QUADAS-2 methodological quality assessment tool (Whiting et al., 2010), which aims to assess the methodological quality/risk of bias of primary diagnostic accuracy studies. The tool consists of four main domains: (1) patient selection, (2) index test, (3) reference standard, and (4) patient flow through the study/time of the index test(s) and reference standard (''flow and time''). The assessment of each study is completed in four phases, and each domain is assessed in relation to the risk of bias (Whiting et al., 2010) (Supplemental 3). The first three domains (patient selection, index test, and reference standard, respectively) are also evaluated with respect to applicability concerns. To help decision making regarding the risks of bias, the instrument has flagging questions. The flagging questions are answered as follows: ''yes'' (low risk of bias/high methodological quality), ''no'' (high risk of bias/low methodological quality) and ''unclear'' (insufficient information to allow for a judgment). If all flagging questions for a given domain are answered ''yes'', the article's risk of bias is considered ''low'', whereas if any question is answered ''no'', the article is considered to have some potential risk of bias. The answer ''unclear'' is only assigned to the item evaluated when there is not enough information to make a judgment. Also in relation to the assessment, although the risk of bias/methodological quality analysis tool does not have flagging questions for attributing judgment regarding the applicability of the strategy adopted by the study under analysis, the authors are requested to record information upon which the applicability judgment is performed and then classify the reason why the study met or did not meet such criteria (Whiting et al., 2010).
This instrument for assessing the risk of bias/methodological quality does not use a ''quality score'' classification, since the analysis of studies occurs in a segmented manner (according to items previously described). However, if a study was judged to be ''low'' in one or more domains related to bias or applicability, then it is appropriate to have an overall judgment of ''low risk of bias'' or ''little concern about applicability'' for that study. If a study was considered ''high'' or ''not clear'' in one or more domains, it could be considered ''at risk of bias'' or ''with concerns about applicability'' (Whiting, Harbord & Kleijnen, 2005).
Eight studies used the 90th or 95th SBP and/or DBP percentiles to classify individuals with HBP, according to sex and age (Beck, Da Silva Lopes & Pitanga, 2011;De Moraes & Da Veiga, 2014;Vogt Cureau & Reichert, 2013;Taylor & Hergenroeder, 2011;Febriana, Nurani & Julia, 2015;Kruger et al., 2013;Mazicioglu et al., 2010). One study followed the guidelines (Roccella, 1996) for classifying pressure levels determined according to age, sex and adjusted height . Thus, in the aforementioned study , if the mean of three measurements of SBP or DBP was greater than the 90th percentile and less than the 95th percentile, adolescents were classified as prehypertensive, if the mean of SBP or DBP was greater than the 95th percentile, adolescents were classified as hypertensive, and if the mean of the three measurements of SBP or DBP was less than the 90th percentile, adolescents were classified as normal blood pressure. In addition, adolescents with SBP > 120 mmHg and DBP > 80 mmHg, but with a percentile < 95, were considered prehypertensive . In another study (Al-Bachir & Bakir, 2017), values of SBP > 135 mmHg and DBP > 89 mmHg were adopted to classify adolescents with HBP. Furthermore, eight studies indicated that anthropometric indicators of obesity had an acceptable predictive capacity to identify HBP in adolescents (Beck, Da Silva Lopes & Pitanga, 2011;Vogt Cureau & Reichert, 2013;Taylor & Hergenroeder, 2011;Abbaszadeh et al., 2017;Al-Bachir & Bakir, 2017;Febriana, Nurani & Julia, 2015;Mazicioglu et al., 2010). However, only in one study (Beck, Da Silva Lopes & Pitanga, 2011) Table 2).
This review identified high risk of bias with regard to subject selection and measurement procedures for anthropometric indicators of obesity in seven studies (Beck, Da Silva Lopes & Pitanga, 2011;De Moraes & Da Veiga, 2014;Vogt Cureau & Reichert, 2013;Taylor & Hergenroeder, 2011;Abbaszadeh et al., 2017;Kruger et al., 2013;Mazicioglu et al., 2010), and high risk of bias in the measurement of blood pressure levels in five studies (Beck, Da Silva Lopes & Pitanga, 2011;De Moraes & Da Veiga, 2014;Vogt Cureau & Reichert, 2013;Kruger et al., 2013;Mazicioglu et al., 2010). Furthermore, one study did not make it clear how procedures to measure blood pressure levels were performed (Taylor & Hergenroeder, 2011). Another aspect that deserves attention is that although eight studies (Beck, Da Silva Lopes & Pitanga, 2011;Vogt Cureau & Reichert, 2013;Taylor & Hergenroeder, 2011;Abbaszadeh et al., 2017;Al-Bachir & Bakir, 2017;Febriana, Nurani & Julia, 2015;Kruger et al., 2013;Mazicioglu et al., 2010) have concluded that the anthropometric indicators of obesity had good predictive capacity for HBP, only in one study (Beck, Da Silva Lopes & Pitanga, 2011) the AUC values were acceptable for use in a clinical context (Akobeng, 2007;Fischer, Bachman & Jaeschke, 2003). Thus, in addition to high sensitivity and specificity values, the stipulated cut-off points must have high LR+ and low LR-values, in addition to allowing a definitive diagnostic condition for the investigated health outcome, regardless of the estimated prevalence. In this context, the use of diagnostic measures such as PPV and NPV are suggested.
Although the use of anthropometric indicators as diagnostic tools to be used for screening for HBP in adolescents has recognized clinical relevance, ethnicity can play a determining role in growth patterns, and consequently, in the precision and accuracy of anthropometric indicators used. Thus, not considering ethnic variation in the assessment of anthropometric obesity indexes to be used for screening for HBP in adolescents can lead to the identification of inaccurate results. However, even though the summarized evidence has been heterogeneous with respect to the investigated populations (adolescents from different regions of the globe-different ethnicities), the small number of studies impaired proposing results and possible suggestion of thresholds to be adopted for the diagnosis of HBP according to ethnicity.
This review presents strengths and limitations that must be considered. The large number of databases investigated (five) in order to identify evidence related to the topic of interest is considered a strong point of this review. Additionally, the analysis of studies in three different languages is another strong point of this review. Despite the careful search carried out by the researchers, it is possible that studies related to the theme covered in this review have not been identified, which is considered a limitation. Thus, the search in a greater number of databases and in the gray literature (for example, course conclusion works and specialization monographs) could contribute to the fact that possible information regarding the theme is not left out of the study. Moreover, the systematic search in other languages could also contribute to bringing new evidence on the investigated topic.

CONCLUSION
Based on the summarized information, caution is suggested in the use of anthropometric indicators of obesity in the HBP screening in adolescents. Although there were studies that have suggested the use of these indicators, these suggestions were not based on AUC measures with high predictive capacity. Thus, further studies are needed that report a high predictive capacity of anthropometric indicators of obesity for the HBP screening.