After removing duplicates, a total of 17543 records were screened, out of which 77 full-text articles were reviewed (see figure 1.0). A total of 27 studies (including 3 studies identified during the update search) met the inclusion criteria. Other full-text studies were excluded because of the following: predicting genetics (n=1), investigating one or more single risk factors which are not connected to build a model or score (n=22), used unconventional predictors (complimentary/alternative medicine) (n=3), applied other disease model or score to predict MetS (CVD, T2DM) (n=4) main outcome is not MetS (n=8) predicts MetS risk reduction (n=1), did not report any related predictive outcome (either discrimination or calibration) (n=9) conducted in languages other than English (n=2).
Furthermore, out of the 27 included papers, 25 report about the development of one or more risk model or score [19-30,32,33,35-45], and 2 studies report about the development and validation of one or more risk model or scores on an external population [31,34]. Overall, the 27 studies reported 43 models, out of which 27 models were selected for full data extraction. The rest (16 models) were not selected, either because they were judged to be minimally different from the reported ones or they were not the preferred models by the authors or they were significantly deficient in details or statistical reporting.
Table 1.0 provides a summary of the quality assessment of the included studies. In summary, the quality assessment revealed that in the entire included studies there is moderate-to-high risk of bias, primarily due to the use of inappropriate study design and absence of external validation. Further look at the studies, it was observed that majority of the models suffered a high risk of bias and significant methodological deficiencies arising from poor choice of model analyses, significantly underpowered analyses, dichotomisation of continuous variables, lack of adjustment for optimisation, poor handling of missing data and overall poor model presentation.
Table 2.0 provides the summary of the included studies. In summary, there is high heterogeneity in the studies. Studies were conducted in 17 countries with sample size ranging from 62 to 36944 participants. Overall, risk models were tested on 155,027 individuals. Furthermore, studies differ in terms of the target population (children/ young vs adults) as well as purpose (diagnosis vs prognosis of MetS). Due to the heterogeneity of data, difference in methodological approach and presentations, it is challenging to make comparisons across studies.
MetS risk models or scores for children
Regarding the target population, more risk scores were reported in adults than in children or adolescents. Of the 27 included models, 17 focused on adult subjects [19,20,24,26-33,36-39,42,45], while 10 targets children and adolescents [21-23,25,34,35,40,41,43,44].
The main purpose of developing most MetS models targeting children and adolescents is diagnosing MetS. Only one model was built for the purpose of predicting MetS prospectively [34]. Similarly, most of the children MetS models were developed using cross-sectional data [21-23,25,35,40,41,43,44] and they often report incidence/prevalence of MetS at the end of the study, which ranges from as little as 1.2% [41] to as high as 54.8% [44]. In all the 10 risk scores aimed at children, various combinations of risk factors (predictors) were considered significant in the respective final model. The average number of predictors utilised in a single risk score is five. Similarly, majority of the studies used a continuous criteria in the diagnosis of MetS [21-23,25,35,40,41,43,44] (see table 3.0).
Poor reporting quality is observed across the studies, especially concerning statistical data. For instance, only one study reported calibration (of any statistic form) [34] and two studies reported positive and negative predictive values [22,34]. The commonest measures reported are sensitivity/specificity (all but one study) [41] and AUROC (all studies). AUROC ranged from 0.87 to 0.98. In the same vein, the final model equation was reported in only 5 studies [21-23,25,41]. Also, one study reported having an online risk calculator [40].
Regarding the validity of the risk models or scores, any form of internal validation is either not conducted, or not reported in nearly a fifth of the studies. Similarly, only one risk model was externally validated [39]; by same authors and, reported in the same paper (with corresponding model development). However, there is some form of uniformity when it comes to the biomarkers used to capture the assumption of MetS with majority of studies using a combination of WC, MAP, HDL-C, TG and FBG. However, 3 studies used HOMA-IR instead of FBG [21,40,44], while a study employed salivary biomarkers (as against blood biomarkers) [41],
MetS risk models or scores for adults
Both diagnostic and prognostic MetS models exist for adults. But, majority (70.6%) are diagnostic developed using data from cross-sectional studies [19,20,24,26,27,31-33,37,39,45]. Surprisingly, the observed incidence/prevalence of MetS at the end of the study is not reported in over fourty-one percent of the studies [19,20,26,32,36,37]. For those that reported it, the prevalence/ incidence of MetS ranged from as little as 7.9% [31] to as high as 57.8% [45]. In all the 17 risk scores targeting adults, various combinations of risk factors (predictors) were considered significant in the respective final model. Again, different weights were assigned to different components in the various models. The number of predictors utilised in a single risk score ranged from 2 to 11 (mean 5.6, SD 1.95). Similarly, NCEP including its modified versions were by far the most widely used criteria (32.5%) in the diagnosis of MetS (see table 3.0).
Similar to the children studies, there is generally poor reporting quality across the studies, especially concerning statistical data. Only two studies reported calibration (of any statistic form) [30,42]. Similarly, more than one in five studies did not report sensitivity/specificity[19,29,30,36]; four-fifth did not report positive and negative predictive values[19,20,24,26-30,32,33,36,39,42]. Similarly, Area Under Receiver Operating Curve (AUROC) ranged from 63.0 to 95.0. One study did not report any discrimination measure [27].
Regarding the validity of the risk models or scores, any form of internal validation is either not conducted, or not reported in nearly a fifth of the studies. Similarly, only one risk model was externally validated [31]. More so, the external validation was done by the same authors and, reported in the same paper with corresponding model development. In addition to the traditional biomarkers/ predictors of MetS (i.e. abdominal obesity, blood pressure, blood glucose, triglyceride and HDL-cholesterol), some studies employed other (novel) biomarkers, such as phenotypic biomarkers (double chin, buffalo hump) [32],Quadriceps muscle peak torque/body mass, (Nm/kg) [33], and lifestyle factors (alcohol, moderate physical activity, smoking, food insecurity, habit of eating less salt, dairy consumption) [31].
Reporting the model equation can add to the reporting quality and replicability of the study. However, the final model equation was reported in only 10 studies [19,20,24,26,27,32,36,37,39,42]. Also, one study reported having a risk estimation chart [42] and another has mobile application and excel risk calculators [19].
Diagnostic vs prognostic MetS risk models or scores
The diagnostic and prognostic MetS models differ in certain characteristics. On the one hand, MetS diagnostic models are usually developed using data from cross-sectional studies, or routine data that were often assembled for other reasons [19-27,31-33,35,37,39-41,43-45]; have smaller sample size; and mostly targeted at young individuals. On the other hand, prognostic MetS models tend to be developed using cohort data (both general population and primary care) [28,29,30,34,36,38,42]; with relatively larger sample size; target mainly adults [28,29,30,34,36,38,42]; have prediction horizon ranging from 2 years [42] to 10 years [34] (see table 3.0).