This systematic review revealed that there are numerous MetS prediction models or scores in the literature. This finding is similar to what is seen in other chronic diseases such as CVD [52,53,54] and T2DM [22,55,56]. Similarly, studies of MetS risk models vary significantly in terms of methodological approach, presentation, purpose (diagnosis vs prognosis of MetS) and target population (children vs adult). However, despite their variations, they share common characteristics.
Poor conduct and reporting of prediction model studies is a common finding reported across most similar reviews [22,52-56], and often, this leads to missing of vital information. This finding is also observed in most of the studies included in this review. The lack of standardised guideline both for conduct and reporting of prediction model studies is believed to be largely responsible for this occurrence [15]. However, with tools such as TRIPOD [57] and PROBAST [23] being developed and validated, the situation is likely to change in the nearest future.
Assessing the overall performance of prediction models is necessary before translating research findings into the real-world setting [58]. However, model performance can be affected in several ways. One of these ways is how continuous variables are handled; are they retained as continuous measures or are they categorised into two or more categories [59]. Categorisation of variables is frequently observed in risk prediction model studies [60-63]. In this review, categorisation of some, or all variables was conducted in (65%) of the studies. This finding is in keeping with what is reported previously in similar reviews [52,54,55]. However, it is recommended that while developing a model, continuous risk variables (predictors) should be retained as continuous variables, or rather, splines or fractional polynomial functions should be used if the relationship between the predictor and the outcome is nonlinear [64].
Another way that the performance of models is affected is through missing values. Missing values are a common occurrence in most datasets [55]. Nearly half of the studies in this review failed to report information regarding how they treated missing values. This finding is in keeping with other similar reviews [52,55,60,63]. Potent ways of minimising the effect of missing values such as the multiple imputation technique have been recommended [65,66]. Therefore, researchers should always report the completeness of the overall data and how the missing values are dealt with so that the representativeness and quality of the data can be judged by readers.
Again, there is a lack of consistency in studies of MetS prediction as they used different predictors and statistical methods. To the very least, discrimination and calibration measures are recommended to be reported [67]. Although nearly all studies reported some form of discrimination, however, calibration is rarely reported. In this review, only three studies reported any form of calibration measure [36,40,48]. This is similar to other relevant reviews [52,55,60,63]. This makes it difficult to make comparison across studies, e.g. meta-analysis and to assess the generalizability of the studies [68].
Furthermore, majority of the studies used common biomarkers (blood pressure, fasting blood sugar, cholesterol, triglycerides and waist circumference) as predictors in building their models. In addition to these, however, other novel predictors /biomarkers have been used once or twice by some researchers. However, none of the models that reported using novel biomarkers has been used elsewhere or externally validated. A similar observation is made in CVD models studies [52]. This shows that researchers in the field give more significance to the process of identifying new predictors and new model building as against validating and applying existing ones.
Regarding the definition/ criteria of MetS, there is significant heterogeneity amongst the studies. But, the NCEP criteria [69] or its modified versions are the most commonly used. This further makes it difficult to compare between studies because different definitions of outcome result in difference in predictor effect and resultant model performance [52]. Having a more uniform definition/ criteria would help significantly in mitigating the above (thereby making it easy to compare between studies and eventually translate research findings into clinical setting) [57].
Furthermore, most of the studies included in this review described developing MetS prediction models, but, the external validation of such models is seriously lacking. Certainly, the ultimate aim of any multivariable model study is to show that the model in question works [55]. It is, therefore; of paramount importance that the model performance is assessed once it is developed [68,70]. Only two models in this review were externally validated. Lack of external validations is a common problem of most prediction model studies [52,55,60,63].
Head-to-head comparison of models assists in knowing which models are better in terms of performance. In this review, no such comparative study is observed. This makes it difficult to choose, or advocate amongst the existing studies. Comparative studies (preferably of multiple models in a single study) are recommended in prognostic risk prediction models [52]. However, as significant as the statistical characteristics of a prediction model may be, they do not guarantee its usefulness in a clinical/ real-life setting. None of the models in this review is reported to have been applied in clinical setting. Therefore, in the future, more emphasis should be given on impact studies- applying the models in clinical setting and assessing their ability to influence decision making or patients’ outcome.
The number of risk models or scores included in the final sample of this review is relatively high. But, we believe none of the existing MetS models or scores in their current state warrants to be applied in real-life setting without being investigated further. For any prediction model or risk score to be considered useful, it should be accurate (statistically significant calibration, and discrimination above 0.70), generalisable (externally validated by a separate research team on a different population) and usable (has few components that are commonly used in practical setting) [71]. However, MetS prediction discipline is arguably still in its early phase of development; therefore, it is difficult to identify any model or score that fulfils all of the above criteria. Nevertheless, for clarity, it was decided to highlight the risk models or scores we feel should be researched further. To aid in making this decision, the above criteria proposed by Altman et al. [71] was modified. In addition, we also considered the availability of either full model equation, online risk calculator or risk estimate chart in the paper, without which it will not be possible to replicate the study. Based on the above, the following prediction models are recommend to be further tested: For the purpose of long term prediction of MetS in adolescence, we recommend the risk model described in Efstathiou et al. [40]. The model was developed using a cohort data, used few predictors and reported good accuracy (discrimination 0.97 & calibration 6.64). Furthermore, the model has been externally validated. Similarly, for the purpose of predicting risk of developing MetS around mid-life, the model developed by Hsiao et al. [48] should be researched further. The model was developed using a primary care cohort, used few number of common predictors, and reported fairly good performance measures (discrimination 0.77 & calibration 0.20). In addition, both the full model equation as well as the corresponding risk estimation chart are provided, making it easy to replicate. Finally, if a researcher is interested in a risk score for self-assessment by older lay persons, then we suggest the JAMRISC [33]. The score was developed using non-laboratory predictors. Also, the developed questionnaire as well as the final model equation are provided which could be used.
The role of MetS as an effective tool for identifying individuals at increased risk of CVD and T2DM has been recognised [72]. Focusing on identification of individuals at risk of MetS will help in the wider prevention of T2DM, CVD and other NCDs related to MetS. However, despite the clear evidence linking it with the risk of CVD and T2DM, the suitability and practicality of applying MetS in clinical setting is still debatable [7]. The major argument for the use of MetS in clinical practice is that physicians can use its presence to improve the care they provide or the outcome of their patients [8].
Generally, abdominal obesity is presumed to be central in the diagnosis of MetS, but, not all individuals with MetS are obese and vice versa [9]. The consideration of metabolically healthy obese and nonobese MetS is gaining more attention [73]. This is even more relevant with the recognition of other significant risk factors associated with MetS such as stress, poverty and unhealthy diet [74-76]. Distinguishing between these groups of MetS might be necessary because of the potential management implications [11]. But, without a validated special test, it may not be possible to detect or suspect MetS in nonobese individuals and those patients who may not be otherwise suspected to have it [10]. Individuals with nonobese MetS may have relatively low perception of risk of MetS, and may be missed by clinicians (because they may not be considered high risk) [10]. Similarly, in obese patients, MetS can be used as a marker indicating increase need of intervention to reduce subsequent risk of CVD and T2DM.
Through the use of MetS risk models or scores we can identify and focus on individuals with the most risk of MetS without necessarily screening the whole population [77]. As shown in this review, MetS risk models could be deployed for different purposes and various populations. The diagnostic MetS models are typically built using common clinical biomarkers which are also components of MetS. Therefore, they could be utilised in everyday clinical setting for assessing high risk group (patients highly suspected to have MetS either based on their clinical presentation or significant history) [78]. Arguably, the diagnosis of MetS may motivate patients and their clinicians to take the necessary actions required for reducing their risk of CVD and T2DM [78]. Similarly, some of the prognostic MetS models could be applied in general prevention/screening programmes targeting low risk group (e.g. apparently healthy young people). Also, they could be used to monitor response to treatment in patients diagnosed with MetS and undergoing therapy.
Identifying and managing MetS is a form of primary or early secondary prevention of CVD and T2DM. Thus, primary care is the most appropriate setting where MetS prediction models should be applied. More so, most of the predictors of the existing MetS models could easily be obtained in such setting. Furthermore, many similar risk models are being utilised in the primary care setting.
Both CVD and T2DM share specific risk factors and are the target of many prevention guidelines [17]. But, so far, no perfect risk model for CVD and T2DM exists. Since the purpose of prevention programmes is to address specific issues before they progress into chronic disease, MetS may become an important assessment tool [11]. The WHO stress that MetS should be viewed as a pre-morbid state of CVD and T2DM, not a tool for diagnosing a distinct clinical disease [72]. Indeed, diagnosing MetS combines several risk factors thereby providing a more holistic picture to the physicians regarding the individual’s risk of CVD and T2DM which is at the core of many primary healthcare systems [11]. However, before this could be achieved, MetS prediction models must be shown to have good discriminative power both in high risk group and general population [56]. Therefore, as a next step, some of the promising MetS models need to be further investigated to ascertain their real worth.
What is already known on the topic
- The findings of this review are in keeping with earlier published reviews of risk prediction models or scores in CVD, T2DM, cancer and stroke.
- Numerous MetS prediction models or scores exist in the literature
- No detailed review has described how these models or scores were developed, their predictive performance and how many were validated (externally).
What this study adds
- Though there are many MetS prediction models or scores, only a few have been externally validated.
- Poor reporting was observed in all aspects of risk prediction models development, specifically, in terms of data description and sufficient detail in all steps taken in building the model.
- This makes them of doubtful value to potential users (practitioners, policy makers and guideline developers).
Strengths
The strengths of the systematic review include:
- This review consists of models or scores that are geographically diverse (across different countries). This is likely to give a broader picture of MetS risk prediction because it has been shown that models performances are affected by race/ethnicity.
- The review used standard methods that apply to reviews of any risk models and scores.
Limitations
This review was limited to articles published in English; therefore; some significant additional findings might be missed. Notwithstanding, it is strongly suspected that none of the study findings would be altered by adding more articles. Furthermore, large numbers of titles and abstracts were screened at the initial stages of the review due to lack of a sensitive literature search strategy. Finally, due to significant heterogeneity between the included studies, detailed quantitative analysis and formal meta-analysis could not be performed.