Development and Validation of Two Self-Reported Tools for Insulin Resistance and Hypertension Risk Assessment in A European Cohort: The Feel4Diabetes-Study.

Early identification of type 2 diabetes mellitus (T2DM) and hypertension (HTN) risk may improve prevention and promote public health. Implementation of self-reported scores for risk assessment provides an alternative cost-effective tool. The study aimed to develop and validate two easy-to-apply screening tools identifying high-risk individuals for insulin resistance (IR) and HTN in a European cohort. Sociodemographic, lifestyle, anthropometric and clinical data obtained from 1581 and 1350 adults (baseline data from the Feel4Diabetes-study) were used for the European IR and the European HTN risk assessment index respectively. Body mass index, waist circumference, sex, age, breakfast consumption, alcohol, legumes and sugary drinks intake, physical activity and sedentary behavior were significantly correlated with Homeostatic Model Assessment of IR (HOMA-IR) and/or HTN and incorporated in the two models. For the IR index, the Area Under the Curve (AUC), sensitivity and specificity for identifying individuals above the 75th and 95th of HOMA-IR percentiles were 0.768 (95%CI: 0.721-0.815), 0.720 and 0.691 and 0.828 (95%CI: 0.766-0.890), 0.696 and 0.778 respectively. For the HTN index, the AUC, sensitivity and specificity were 0.778 (95%CI: 0.680-0.876), 0.667 and 0.797. The developed risk assessment tools are easy-to-apply, valid, and low-cost, identifying European adults at high risk for developing T2DM or having HTN.


Ethics Approval and Consent to Participate
The Feel4Diabetes-study adhered to the Declaration of Helsinki and the conventions of the Council of Europe on human rights and biomedicine [25]. All participating countries obtained ethical clearance from the relevant ethical committees and local authorities. More specifically, in Belgium the study was approved by the Medical Ethics Committee of the Ghent University Hospital

Study Protocol and Recruitment
A detailed description of the methodology of the Feel4Diabetes-study has been previously published [26]. Briefly, the recruitment of the population study was carried out via a multi-stage sampling procedure in selected provinces of six European countries (Hungary, Bulgaria, Finland, Belgium, Greece and Spain). In each country, primary schools located in the selected municipalities were used as the entry-point to the community and parents having children attending the first three grades were invited to complete the Finnish Diabetes Risk Score (FINDRISC) and a brief questionnaire on lifestyle habits (self-reported data). If at least one parent fulfilled the country-specific cut-off point for FINDRISC (for the majority of countries that was set as a FINDRISC score ≥ 9), both parents (regardless their individual FINDRISC score) were invited to undergo a brief medical check-up. This procedure led to a cohort with a wide distribution of FINDRISC values among the participating parents, i.e., 25.9%, 37.3% and 36.8% of participants had a total FINDRISC score <9, 9-11 and ≥12 respectively.
Exclusion criteria for the development of the two risk assessment indices were: previous diabetes diagnosis, not following the fasting protocol, current antihypertensive treatment and incomplete data. Therefore, from the initial 3153 parents, 1572 were excluded from the Homeostatic Model Assessment of IR (HOMA-IR) index and 1786 were excluded from the HTN index, and this resulted in 1581 (Bulgaria, n = 367; Finland, n = 218; Belgium, n = 214; Greece, n = 476; and Spain, n = 306) and 1350 (Hungary, n = 19; Bulgaria, n = 361; Finland, n = 278; Belgium, n = 173; and Greece, n = 519) subjects for each model, respectively. Insulin data were not analyzed in Hungary, thus the data obtained from this country was not included the IR risk index. Similarly, the data obtained from Spain was excluded from the HTN risk index, as alcohol intake was not recorded in this cohort. Blood pressure and blood indices (i.e., insulin and glucose) were measured by trained researchers using standardized procedures and calibrated equipment in every country as described elsewhere [26]. All variables used in the two developed tools were self-reported (including anthropometric measurements).

Measures
Questionnaire data: All the relevant data, such as sociodemographics (i.e., sex, age, educational level, marital status, etc.), behavioral indices regarding dietary habits, physical activity and sedentary behaviors (i.e., portions of sugary drinks per week, number of meals and snacks during a day, minutes of daily vigorous physical activity, time spent in front of computers and television, etc.) were collected from participants [26].
Anthropometry: All study participants received paper measuring tapes and brief written instructions on how to measure their height, weight and waist circumference. BMI and waist circumference were classified based on the World Health Organization (WHO) criteria [27].
BP measurement: BP was measured on the right arm, in a sitting position using electronic sphygmomanometers (OMRON M6 or OMRON M6 AC, Omron Healthcare, Kyoto, Japan) after five minutes of rest, on three occasions, at one-minute intervals. Participants were classified according to the European guidelines [28] in the following categories: optimal (systolic BP < 120 mmHg and/or diastolic BP < 80 mmHg), normal (systolic BP 120-129 mmHg and/or diastolic BP 80-84 mmHg), high normal (systolic BP 130-139 mmHg and/ or diastolic BP 85-89 mmHg), grade 1 HTN(systolic BP 140-159 mmHg and/ or diastolic BP 90-99 mmHg), grade 2 HTN (systolic BP 160-179 mmHg and/or diastolic BP 100-109 mmHg), and grade 3 HTN (systolic BP ≥ 180 mmHg and/or diastolic BP ≥ 110 mmHg), with the BP category to be defined by the highest level of BP, either systolic or diastolic.
Blood indices: Blood samples were drawn in the morning after overnight fasting (duration: eight hours or longer). FPG and fasting insulin was analyzed in accredited laboratories, using similar enzymatic assays in all study centers. HOMA-IR was calculated as indicated by Matthews et al. [29].

Statistical Analysis
Factors associated with HOMA-IR and HTN were identified by Pearson's or Spearman's correlation and analyses of variance. For dietary behavior and physical activity factors, receiver operating characteristic (ROC) analysis was performed and the optimal cut-offs were determined by the point with the shortest distance to (0,1) in the ROC curve that maximizes the sensitivity (Se) and specificity (Sp) of the test. The distance for each observed cut-off was calculated as the square root of [(1 − Se) 2 + (1 − Sp) 2 ] [30]. For other variables, such as waist circumference and BMI the cut-off values set by WHO were used. The population was randomly separated to 2/3 and 1/3 for the development (n = 1076 IR cohort and n = 906 HTN cohort) and validation (n = 505 IR cohort and n = 444 HTN cohort) of the indices respectively. For the development of the risk assessment indices, backward linear regressions were performed, with dependent variables the percentiles of HOMA-IR and 5 categories of HTN classification, respectively. The exclusion criteria of independent variables were set at p > 0.10. The adjusted β-coefficients were multiplied and rounded to the nearest integer value as needed so as to sum up to 40 points. In order to validate the indices, ROC analysis was performed to the validation cohort. The scores with the best combination of Se and Sp (determined as described previously), for 75th and 95th percentile for HOMA-IR, and for grade 2 and 3 HTN, respectively, were used as cut-off scores for the interpretation of the results. The statistical analyses were performed using the Statistical Package for Social Sciences (SPSS Inc., Chicago, IL, USA), version 21.0.

Results
The descriptive characteristics of the study population for the European IR Risk Index and the European HTN Risk Index are presented in Table 1. Moreover, the sex distribution was 33.8% males and 66.2% females and 31.9% males and 68.1% females, for European IR Risk Index and the European HTN Risk Index, respectively. None of the aforementioned variables differed significantly between development and validation cohorts. Regarding the European IR Risk index, the parameters found to be significantly associated with the HOMA-IR were: number of breakfast occasions per week, unsweetened and sweetened milk consumption, sugary drinks consumption, fish, red meat, fruits and vegetables consumption, BMI, sex, waist circumference, number of walking sessions during the week lasting at least 30 min, number of vigorous physical activity sessions during the week lasting at least 10 min and leisure screen time (i.e., television, video games, computers, etc.). At the final model the variables found to be statistically significant in identifying the HOMA-IR percentiles after the stepwise procedure were BMI, screen time, sex, breakfast, sugary drinks, waist circumference, walking and vigorous physical activity as shown in Table 2. As for the European HTN Risk Index the parameters found to be significantly associated with the BP categories as determined by the ESH were: number of cigarettes per day, number of lunch occasions per week, morning snacks per week, afternoon snacks per week, evening snacks per week, whole grain foods consumption (i.e., whole wheat bread, breakfast cereals), fruits and vegetables, red meat, white meat, fish, nuts and legumes consumption, alcohol intake, and leisure screen time (i.e., television, video games, computers, etc.). At the final model the variables that were found to be statistically significant in identifying grade 2 and 3 HTN after the stepwise procedure were legumes consumption, alcohol intake, sex, age, BMI, and vigorous physical activity, as shown in Table 3. The ROC analysis in the validation cohort indicated an area under the curve (AUC) 0.768 (95%CI: 0.721-0.815) for identifying individuals above the 75th percentile of HOMA-IR. The index cut-off score for identifying individuals above the 75th percentile was 23/40, as indicated by the optimal match of Se and Sp (0.720 and 0.691, respectively) (Table 4, Figure 1). Furthermore, the indicated AUC for identifying individuals above the 95th percentile of HOMA-IR was 0.828 (95%CI: 0.766-0.890). The index cut-off score for identifying individuals above the 95th percentile was 31/40, as indicated by the optimal match of Se and Sp (0.696 and 0.778, respectively) (Table 4, Figure 2). Regarding HTN, the ROC analysis in the validation cohort indicated an AUC 0.778 (95%CI: 0.680-0.876) for identifying individuals with grade 2 and 3 hypertension. The index cut-off score for identifying individuals with grade 2 and 3 hypertension was 26/40, as indicated by the optimal match of Se and Sp (0.667 and 0.797, respectively) (Table 4, Figure 3).

Discussion
The ongoing increase in the prevalence of T2DM and HTN and their acknowledged impact on public health highlight the strain for more appropriate strategies and tools for their early diagnosis and prevention. Since both of these metabolic abnormalities may be provoked by the interaction of various risk factors relevant to family history, anthropometric indices and lifestyle parameters, such

Discussion
The ongoing increase in the prevalence of T2DM and HTN and their acknowledged impact on public health highlight the strain for more appropriate strategies and tools for their early diagnosis and prevention. Since both of these metabolic abnormalities may be provoked by the interaction of various risk factors relevant to family history, anthropometric indices and lifestyle parameters, such as dietary behaviors and physical activity, their assessment is necessary so as to design effective prevention strategies. Therefore, the development and implementation of screening tools evaluating holistically various health-related variables can be of a great importance in order to easily identify

Discussion
The ongoing increase in the prevalence of T2DM and HTN and their acknowledged impact on public health highlight the strain for more appropriate strategies and tools for their early diagnosis and Nutrients 2020, 12, 960 9 of 13 prevention. Since both of these metabolic abnormalities may be provoked by the interaction of various risk factors relevant to family history, anthropometric indices and lifestyle parameters, such as dietary behaviors and physical activity, their assessment is necessary so as to design effective prevention strategies. Therefore, the development and implementation of screening tools evaluating holistically various health-related variables can be of a great importance in order to easily identify people at risk, so as to be referred for further evaluation at primary healthcare settings. The aim of the current study was to develop two risk assessment indices for the identification of IR and grade 2 and 3 HTN.
A number of risk scores to predict T2DM risk already exist in the literature [14,15,19,31], but most of them do not apply at various cases such as different races or socioeconomic status. One of the most reliable and valid tools for the identification of people at increased risk for T2DM is the FINDRISC [13], which was initially developed on the Finnish population, showing high Se and Sp values. Thereafter, it has been validated in other European ethnicities with good validity results [32][33][34][35]. Despite the fact that the FINDRISC has been designed on a specific population, it seems that the use of different cut-offs in different national groups may manage equivalent accuracy [36][37][38]. Similarly, an accurate easy-to-use online tool for the detection of impaired glucose regulation and T2DM is the Leicester Risk Assessment Score, an index developed in a multiethnic UK population with a high probability to identify population at risk [18]. Likewise, the Australian index AUSDRISK (The Australian Type 2 Diabetes Risk Assessment Tool) is considered of satisfactory Se and Sp but no studies have examined its validity in different ethnicities [14]. In general, the majority of tools available in the literature have been found to have satisfactory predictive values, but these values may be limited in the country developed [31] or some score components that are prerequisite for the completion of some tools such as blood indices measurements or the awareness of unfavorable glycaemia indices [13,15,19] makes them non-applicable when individuals have not previously undergone any blood testing.
The aforementioned tools are used for the assessment of the T2DM risk, the early identification of which is a major public health issue. Indeed, the early identification and management of IR may significantly decrease the risk for T2DM and cardiovascular disease (CVD) [39]. Furthermore, individuals in the highest HOMA-IR quintiles for IR have even greater risk for CVD [40]. The main benefit of pre-diabetes early identification is that it does not require medical management. The modification of lifestyle parameters, such as increase physical activity or decrease time devoted to sedentary activities, weight loss, adopting healthy dietary habits, etc., may be important measures in order to prevent the onset of glycaemic abnormalities [41][42][43][44], and thus to lower its economic public health burden [45].
Likewise, the existing risk scores for predicting HTN are either developed in population-specific groups or use as variables the measurements of systolic and diastolic BP [16], which has a large impact on the feasibility and thus the reach of these screening tools. For instance, the population cohort in the risk score of Kshirsagar et al. was middle-aged and older adults [17], and despite a satisfactory AUC (0.75-0.78), it is difficult to apply it in different age groups. Similarly, the risk index of the Strong Heart Study was developed in an American Indian population [23]. Despite the fact that obesity and hyperinsulinaemia affect less the BP of American Indians [46], it seems that the development of HTN in this ethnic group has a more severe effect on cardiovascular health [47]. One of the possible reasons for these findings in the American Indians is the presence of three genes with multiple single nucleotide polymorphisms associated with systolic BP [48]. Therefore, the development of more population-representative and self-reported risk scores is needed to identify HTN accurately.
In the current study, we developed two self-reported risk assessment indices, the European IR Risk Index and European HTN Risk Index for the identification of IR at 75th and 95th percentile according to HOMA-IR score and grade 2 and 3HTN. Both risk scores are calculated from a total of eleven components, and no biochemical, BP or other measurements are required. This makes these indices very easy-to-calculate, and applicable for a wide range of populations regardless their access to medical equipment, healthcare services, or their willingness to go through a medical check-up. Furthermore, by including anthropometric, dietary and physical activity components in the indices, the importance of lifestyle modification is highlighted. Specifically, the importance of increasing physical activity, along with breakfast and legume consumption, and of decreasing sugary drinks and alcohol consumption is highlighted. In addition, based on the ROC analyses, both the European IR Risk Index and the European HTN Risk Index were found to perform well in identifying individuals above 75th and 95th percentile of HOMA-IR and at grade 2 and 3 HTN ( Table 4). The AUCs of 0.768, 0.828, 0.778, respectively, are slightly higher than the relevant of risk assessment indices available in the literature, which range from 0.72 to 0.78 [13,14,17,18,31,49], despite not incorporating biochemical and BP measurements. Another great advantage of the current indices is that the study population in which they have been developed and validated is from six countries in Europe, making them applicable and preferable for a wide range of Caucasian populations. The potential limitations of our study could be the fact that although it is based on a community cohort, the cohorts recruited from each country are not representative of the general population, as the recruitment took place only in one large region within each country, and with probably different risk of developing T2DM than the overall population according to FINDRISC. Overall, the findings are encouraging for the next steps in validating these risk assessment indices in more specific European populations.

Conclusions
T2DM and HTN are, without any doubt, the major risk factors for CVD. Moreover, there is robust evidence that they can be prevented and treated through lifestyle interventions. The disparity among populations around Europe has highlighted the need for simple novel screening tools, without the use of biochemical or other clinical indices in order to identify individuals at increased risk, so as to motivate them to seek medical assistance and proceed with lifestyle modification. The developed European IR Risk Index and European HTN Risk Index in the present study are easy-to-apply, valid, non-invasive, and low-cost for identifying European adults at high risk for developing T2DM or having HTN.