Optimal Cutoff of the TG/HDL-c ratio for Cardiovascular Risk in Hypertensive and Diabetic Patients Monitored by Primary Health Care in a city in Minas Gerais

Background: The analysis of the atherogenic potential of the lipid profile for biomarkers, such as the TG/HDL-c ratio, predicts cardiovascular risk better than isolated lipids. Objective: To identify the TG/HDL-c cutoff points for multiple risks (hypertension, Diabetes Mellitus, obesity) and to evaluate the association between sociodemographic, clinical, laboratory, anthropometric, and life habit variables and the TG/HDL-c ratio in hypertensive and/or diabetic individuals in the context of Primary Health Care. Methods: This was a cross-sectional study with 833 hypertensive and/or diabetic patients, conducted between August 2017 and April 2018. The cutoff point of the TG/HDL-c were obtained by the ROC curve. Cardiovascular risk was discriminated by TG/HDL-c, categorized by the cutoff and evaluated in relation to multiple risks. The magnitude of the association between TG/HDL-c and independent variables was estimated by logistic regression. The significance level of p<0.05 was adopted for all tests. Results: The cutoff values of TG/HDL-c (3.26 for men and 2.72 for women) were more sensitive and less specific than those in the literature. Women (OR=1.90 and 95% CI 1.13-3.20) and men (OR=4.58 and 95% CI 1.78-11.76) with multiple risks, and white men, alcohol users, with a history of stroke, had a higher chance of altered GT/HDL-c. Increases in glycosylated hemoglobin, glycemia, and phosphorus in women, and cholesterol, glycemia, and microalbuminuria in men increased the chances of altered TG/HDL-c. Being a former smoker and black reduced the chance of altered TG/HDL-c in women. Conclusions: TG/HDL-c proved to be a good indicator for habitual use in Primary Care.

18 years of age, who were hypertensive and/or diabetic and who received followed up by the Family Health Strategy. Excluded from the data collection were those individuals who did not continue their follow-up visits, as well as pregnant women, abusive users of alcohol and/or drugs, individuals with severe clinical conditions, and those with established chronic kidney disease (CKD).

Data collection
Data were collected in the Basic Health Units between August 2017 and April 2018. Sociodemographic data, life habits, and health care were collected through semi-structured questionnaires, previously tested in a pilot study, applied by trained researchers. Blood pressure was measured by trained professionals and classified according to the 7 th Brazilian Guidelines on Arterial Hypertension (2016). 14 The weight, in kilograms (Kg), was obtained on an electronic scale with a capacity of 150 kg and division of 50 grams. Stature, in meters, was measured in a portable anthropometer, with a metal platform for positioning of individuals and dismountable wooden column, with millimeter tape and cursor for reading, according to Jelliffe techniques (1966). 15 The body mass index (BMI), calculated by the Weight/Stature 2 ratio (Kg/m 2 ), was classified according to World Health Organization (WHO) criteria (2000) 16 for adults, and Lipschitz criteria (1994) 17 for the elderly. The waist and hip perimeters were measured in centimeters (cm) with inextensible measuring tape. The hip perimeter values were obtained at the level of the maximum extension of the buttocks, with the tape positioned transversely to the measured segment, on the skin, without excessive pressure. Waist perimeter values were obtained at the midpoint between the iliac crest and the external face of the last rib and classified as "increased" in relation to the risk for non-communicable chronic diseases when they presented measurements of ≥94cm for men and of ≥80cm for women, according to the WHO (2000). 16 The waist-hip (WHR) and waist-height (WHT) relationships were calculated by dividing the waist perimeter values by hip perimeter and stature, respectively. The reference values for CVR of the WHR for men (≥0.90) and women (≥0.85) were those recommended by the WHO (2000), 16 while the WHT (≥0.5) between genders was recommended by Ashwell and Hsieh (2005). 18 CVR indices included the Castelli I Indices (CT/HDL-c ratio) and II (LDL-c/HDL-c ratio), 1 and the Framingham score. 1,6 Among the estimates calculated from routine laboratory parameters, covered by Primary Health Care (PHC), the proportion of triglycerides in relation to HDL cholesterol (TG/HDL-c ratio) is easily obtained from the patient's lipid profile. 5,[7][8][9] The TG/HDL-c ratio, proposed by Gaziano et al. 10 as a strong lipid predictor for acute myocardial infarction, has been used as an indicator of dyslipidemias 8 and cardiometabolic risks (obesity, AH and DM), 5,7-9 being a potent predictor of the development of CAD. 8,9,11 The TG/HDL-c ratio dispenses with personnel and specialized techniques, 12 it is a safe, economical, fast-to-obtain, practical, and easy-to-use atherogenic marker. 2,7 For these reasons, its use can be especially considered in PHC. 8 Several studies have suggested cutoff points to indicate CVR, whether equal values for Brazilian elderly men and women 8 or different values for Japanese adult men and women. 5 However, studies whose cutoffs have been calculated specifically for the Brazilian hypertensive and/or diabetic population are unknown. Thus, the present study's objective is to identify the cutoff of the TG/HDL-c ratio for multiple risks (AH, DM and obesity) and to evaluate sociodemographic, clinical, laboratory, anthropometric, and life habit factors associated with the altered TG/HDL-c ratio in individuals diagnosed with AH and/or DM, in the context of PHC.

Study design, sample size calculation, and participants
This is a cross-sectional study that is part of a larger project, 13  Biological samples were collected after 12 hours of fasting, and the biological materials were analyzed in an accredited laboratory, using commercial kits and techniques, together with reference criteria. Microalbuminuria (mg/dL) tests were performed, as were tests for serum albumin, phosphorus, calcium, and creatinine (mg/dL); fasting glucose (FG) (mg/dL); glycosylated hemoglobin (HbA1c) (%); triglycerides (TG) (mg/dL); total cholesterol (TC) and fractionshigh lipoproteins (HDL-c) and low densities (LDL-c) (mg/dL). The TG/HDL-c ratio (dependent variable) was calculated from plasma lipid dosages, dividing the TG values by cholesterol linked to HDL-c. The results of FG and HbA1c were classified as altered (FG-126mg/dL and HbA1c-6.5%) according to the criteria of the American Diabetes Association, adopted by the Brazilian Diabetes Society (2018). 19 The Glomerular Filtration Rate (GFR) (mL/min/1.73m 2 ) was estimated from serum creatinine by the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation. The criterion for CKD was based on Kidney Disease: Improving global Outcomes (KDIGO 2012), 20 considering the values of GFR < 60 mL/min/1.73m 2 .

Statistical analysis
Statistical analyses were performed using computer software programs (SPSS version 20). For descriptive analysis and characterization of the population, absolute and relative frequencies of categorical and mean variables, medians, standard deviations, and interquartile intervals of continuous variables were estimated. Pearson's chi-square test was used to verify associations between categorical variables. The statistically important differences between the continuous variables were verified by the unpaired Student t-tests (parametric) or Mann Whitney U (non-parametric) test, according to the normality of the data. The normality of the distribution was tested by the Shapiro-Wilk test. The significance level of p<0.05 was adopted for all tests.
The cutoff of the TG/HDL-c ratio for CVR discrimination in relation to multiple risk factors (AH, DM, and obesity classified by BMI) were obtained by the Receiver Operating Characteristic (ROC) curve. The ideal cutoff points were selected maximizing the Youden index. The discrimination of the TG/HDL-c ratio was measured by the area under the ROC curve (AUC). The 95% confidence interval (CI) for AUC was estimated by the DeLong method (1988). The sensitivity, specificity, and accuracy of the identified cutoff points, and others already described in the literature, 5,8 were presented.
The magnitude of the association between the TG/HDL-c ratio (categorized by the obtained cutoff points) and the population characteristics were estimated by logistic regression models. The analyses were stratified by sex. Bilateral probability (p) values of less than 0.05 were considered to indicate statistical significance in the multivariate model. To evaluate the magnitude of the associations, the Odds Ratio (OR) and respective 95% CI were used. Table 1 shows the characteristics of the subjects by gender. The participants were classified as overweight by BMI and presented WHR and WHT in the increased CVR range. The median values of BMI and WHT in women and WHR in men were higher. CVR-related variables were significantly higher in women, and 27 men and 87 women accumulated the three risk factors. Figure 1 shows the results of the ROC analysis for the relationships between TG/HDL-c and RCV factors. The optimal cutoff values of the TG/HDL-c ratio for multiple risks were 3.26 for men and 2.72 for women (p<0.001), lower than the reference ratio (=3.5) for both genders, 8 as well as for women (=3.75) 5 and for men (=3.0). 5 The new values showed greater accuracy and sensitivity, and lower specificity than conventional ones ( Table 2).

Results
Considering the cutoff points established by the ROC curve, the TG/HDL-c ratio was categorized as adequate (<3.26 for men, <2.72 for women) and changed (≥3.26 for men, ≥2.72 for women). Men with altered TG/HDL-c ratio were observed as more obese; more frequently of multiple risks; higher median values of weight, BMI, hip perimeter, WHT, and WHR; higher waist perimeter averages; higher mean levels of TG, FG, HbA1c, creatinine, and albumin; and lower HDL-c. A higher proportion of men with adequate TG/HDL-c used hypoglycemic agents (Table 3). Women with an altered TG/HDL-c ratio were observed as more obese, with a higher frequency of multiple risks; median values of weight, BMI, hip perimeter, WHT, and WHR; higher waist perimeter averages; higher median levels of TG and FG; and lower HDL-c when compared to those with an adequate TG/HDL-c ratio. A higher proportion of women      with an altered TG/HDL-c ratio were diabetic, used hypoglycemic agents and were active smokers, while a higher proportion of former smokers and nonsmokers showed an adequate TG/HDL-c ( Table 3).
The reasons for chance (95% CI) for the variables that remained in the final multivariate model were presented by gender (Table 4). Men with multiple risks were 4.58 times more likely to have an altered TG/HDL-c ratio than those without multiple risks, and frequent users of alcoholic beverages were 3.29 times more likely to have an altered TG/HDL-c ratio than non-users. Participants with a previous history of stroke had a 2.90 times higher chance of altered TG/HDL-c ratio than those without this history. A correlation was found between the altered TG/HDL-c ratio and increased TC, FG, and microalbuminuria. White individuals showed a chance of having an altered TG/HDL-c ratio that was 2.40-fold higher than individuals of black ethnicity/color. For women with multiple risks, the chance of altered TG/ HDL-c increased by 90%. Smoking cessation (former smokers) represented a protective factor, decreasing the chance of altered TG/HDL-c by 2.86-fold when compared to active smokers. The chance of altered TG/HDL-c increased by 33% for each 1% increase in HbA1c, by 1% in each 1 mg/dL increase of FG, and by 61% in each 1 mg/ dL increase of phosphorus. Self-declared brown/yellow/ indigenous and white women were twice as likely to have altered TG/HDL-c than black women (Table 4).   Values expressed in absolute numbers (percentages), averages ± standard deviations, medians (percentiles

Discussion
In the present study, the cutoff points for the TG/ HDL-c ratio of 3.26 for men and 2.72 for women, adults, and the elderly, hypertensive and/or diabetic, users of PHC were identified. These results are lower than those used in Brazil in both male and female elderly individuals, 8 those found for Japanese adults, 5 and much lower than conventional cutoff values of 3.75 in men and of 3.00 in women calculated using each of the cutoff values for triglycerides (150 mg/dL in men and women) and HDL cholesterol (40 mg/dL in men and 50 mg/dL in women).
High plasma level of LDL-c and TG, and low levels of HDL-c, are important factors of CVR. 3,21 Lipid reasons can be used for early detection of individual CVR. 5 The LDL-c/HDL-c ratio is a classic index for predicting AD, but the TG/HDL-c ratio is the best predictor for acute myocardial infarction, associated with insulin resistance and metabolic syndrome. 5 The TG/HDL-c ratio correlates directly with plasma LDL-c levels, type B, 5,8 reported as an independent CVR factor. 5 To identify cardiac and metabolic threats, it is important to use different TG/ HDL-c ratio cutoff points between genders, 5,11 as the HDL-c level is higher in women. 5 Different cutoff values for men and women are, in fact, used in the National Cholesterol Education Program's (NCEP) criteria for metabolic syndrome. 22 Therefore, it is reasonable that there is also a gender difference in the cutoff of the TG/ HDL-c ratio: the values were higher in men than in women ( Figure 1). Thus, it is preferable to use different cutoff values of the TG/HDL-c ratio for men and women.
It was evidenced that the altered TG/HDL-c ratio, identified from the cutoff points found in this study, was associated with the presence of multiple risks (AH, DM, and obesity), ethnicity, alcohol use, smoking, history of stroke; CT, FG and HbA1c dosages; and high microalbuminuria and high serum phosphorus. Other studies have identified associations between the TG/HDL-c ratio and several cardiometabolic risk factors, such as alcohol use, 7 smoking, 8 metabolic syndrome, 23,24 oxidative 25 and inflammatory profiles, 12,25 adverse events, 10,26 various anthropometric parameters, 5,8,9,24,27 dyslipidemias, 12,24,27 HA. 5,8,11 DM,5,8,11,24,28 insulin resistance, 11,23,28,29 and renal function. 30 In this sense, the cutoff points found represent the CVR well and are therefore good risk markers for the studied population.
The cutoff values of the TG/HDL-c ratio can be obtained through quartiles; 10 tertiles; 11 for convenience, from values already used by other researchers; 8,9,26,28,30 or even calculated through the ROC curve. 5 Gaziano et al. (1997) 10  The importance of identifying cutoff values by ROC analysis is due to obtaining more satisfactory values for this population (hypertensive and diabetic), which presents more CVR factors, in the instance where it is attended (PHC). The cutoff for the TG/ HDL-c ratio depends on its associated result, and a result that produces greater accuracy in ROC analysis is preferable to determine the cutoff value of the TG/ HDL-c ratio. However, because they are obtained for a specific population, the suggested cutoff cannot be extrapolated to the population in general. However, in a study of cardiometabolic risk factors (AH, DM, and visceral obesity) in periodic health examination records of 10,196 Japanese adults, it was concluded that the power of discrimination of cardiometabolic risk factors of the TG/HDL-c ratio, using conventional cutoff values and obtained by ROC analysis, were similar when applying both methods. 5 The values suggested in the studies mentioned above or presented in the study by Wakabayashi and Daimon (2019) 5 may not fit this population, as they were conducted in other countries and/or with populations with CVRs that were different from those to which hypertensive and/or diabetic patients are subject, or because they were not calculated, but obtained for convenience in the literature. In the present study, the accuracy values for the TG/HDL-c ratio in relation to multiple risks (AH, DM, and obesity) were 0.698 in men and 0.620 in women (Figure 1), which are generally evaluated as low precision (AUC: 0.5 ~ 0.7) but were higher than the accuracy presented by conventional cutoff values (Figure 1). Recent prospective studies conducted in Iran 31 and China 32 showed similar accuracy (0.575 and 0.647).
Considering that the components (TG and HDL-c) are simple, and are already found in routine laboratory tests, 23 the TG/HDL-c ratio can be easily obtained from the patient's lipid profile. 5,[7][8][9] The use of CVR predictors is relevant in clinical practice, 9 and the use of the TG/HDL-c ratio as one of these indices can avoid the indiscriminate use of laboratory tests and related expenses. 23 In addition, the TG/HDL-c ratio has specific characteristics, such as simplicity, low cost, applicability, 27 ease of execution, 5 reliability, practicality, speed in obtaining results, and non-invasive test qualities, 8 making it a useful indicator to predict CVR in routine and screening tests, 5 especially in the context of primary health care. 8

Strengths and limitations of the study
The present study presents as strengths the achievement of the cutoff values of the TG/HDL-c ratio by ROC analysis with a more satisfactory and effective result to discriminate CVR; having been conducted with adults and the elderly -a population with a tendency to present more CVR-factors, and at the level of PHC -aninstance in which hypertensive and diabetic patients are treated. Limitations of the study include a cross-sectional design, which is insufficient to express a causal association between the TG/HDL-c ratio and the studied variables; the difficulty of comparison with other studies due to the methodological differences of obtaining results and the cutoff values of TG/HDL-c; as well as the non-analysis of food intake and physical activity data. It is suggested that longitudinal, multicenter, and/or prospective additional studies should be conducted to discuss the causative relationships and temporal correlations of CVRs with the TG/HDL-c ratio.

Conclusion
For a population of hypertensive and/or diabetic patients, cutoff values for the TG/HDL-c ratio (3.26 for males and 2.72 for females) were lower than those commonly used in clinical practice. These values showed greater accuracy and sensitivity and less specificity than conventional values. It was also observed that the new cutoff points indicative of altered TG/HDL-c were associated with the presence of multiple risks (AH, DM, and Obesity), ethnicity, alcohol use, smoking, history of stroke, and increased values of TC, FG, HbA1c, microalbuminuria, and serum phosphorus.
These results suggest the use of new cutoff points in the clinical practice of follow-up of patients with AH and DM in PHC, aiming to achieve early screening and the appropriate treatment of risk factors that may indicate an undesirable prognosis in this population.