Age- and sex-specific predictive performance of obesity indices for hypertension in the Persian (Shahedieh) cohort study

Abstract

Background

The rising global prevalence of obesity and its associated negative health outcomes impose a substantial burden on healthcare systems. This study aimed to investigate the predictive power of obesity-related indices for hypertension (HTN) in an Iranian population.

Methods

This prospective study included 7793 individuals aged 35–70 years, without HTN, who were enrolled in the Shahedieh cohort study in 2016. During the 4-year follow-up period, HTN status was determined via annual telephone interviews. A conditional time-dependent receiver operating characteristic (ROC) curve was applied to estimate the predictive power of obesity-related indices, including Body mass index (BMI), waist circumference (WC), wrist circumference and Hip circumference, Waist- To- Height Ratio (WHtR), and Waist-to-hip Ratio (WHR) at different ages and separately for each sex.

Results

BMI predicted HTN well in middle-aged and older men, whereas it demonstrated acceptable predictive power in women aged under 60 years. Hip circumference showed acceptable predictive performance for men from middle age onwards, as well as for young and middle-aged women. Although WC could not predict HTN in young and older men, it showed approximately good predictive power for HTN in women across all age groups. WHR did not exhibit sufficient predictive power for HTN in men except at young ages, whereas it demonstrated good predictive power for women of all ages. WHtR and wrist circumference predicted HTN in men from middle age onwards. Similarly, HTN was well predicted in women of all ages except the elderly based on WHtR. The predictive power of wrist circumference fluctuated with age in women.

Conclusions

BMI was the most reliable predictor for men, while for women, waist circumference and WHR performed best across most age groups, and these patterns seemed to hold up well despite the limited number of cases.

Text box 1. Contributions to the literature
• The person's age and sex are strong indicators of their risk for developing hypertension.
• Most of the obesity indices are much better at predicting hypertension in women than in men.
• This research offers a practical roadmap for healthcare workers to recognize patients who are most likely to develop hypertension based on their sex, age, and body type.
• The findings can lead to better personalized health screening by ensuring people get the prevention advice at the right time in their lives.

Background

The prevalence of obesity and its adverse consequences, which are increasing worldwide, are significant burdens to healthcare systems [1, 2]. The World Health Organization (WHO) defines obesity as an “abnormal or excessive accumulation of body fat that may impair health,” and also states that “the fundamental cause of obesity and overweight is an energy imbalance between calories consumed and calories expended” [2]. Obesity is the fifth most common cause of death in the world [2]. Based on the Global Burden of Disease study reports, the prevalence of obesity has more than doubled since 1980 [2]. Obesity is no longer a public health issue restricted to high-income countries, as it has an alarming increase rate due to urbanization, changes in diet, and the adoption of sedentary lifestyles across the developing world over the last two decades [2, 3].

One of the most common comorbid conditions related to obesity is hypertension (HTN), so changes in body weight are coincident with changes in HTN [2, 4]. Epidemiological studies showed that being overweight or obese caused about 65–75% of primary (essential) HTN [2]. Persistent HTN is a factor for stroke, myocardial infarction (MI), heart failure, arterial aneurysm, and chronic kidney failure, and a moderate increase in arterial blood pressure increases the risk of heart disease and shortened life expectancy [4].

Even though HTN is a primary risk factor for cardiovascular diseases leading to increased morbidity and mortality, timely and proper treatment of HTN undoubtedly can extend life expectancy, improve quality of life, and significantly contribute to the success of primary and secondary prevention of cardiovascular disease [5, 6]. Since obesity is a significant modifiable risk factor for HTN and other CVDs, an initial but crucial step to reduce the burden of these diseases is identifying individuals at high risk of HTN and CVD to have an early diagnosis. Obesity-related indices can be used as convenience and economic tools for HTN and other CVD risk screening [7]. The association between obesity and blood pressure is supported by various epidemiological studies, and many of them try to assess the best indicator for hypertension among the obesity-related indices [8]. The strength of this association can vary based on different factors like place of residence (urban or rural), gender, age, ethnicity [7, 9, 10], so the best indicator of HTN remains controversial [8]. Several studies have indicated that the best predictors were BMI and WC among elderly Mexican men and women [11], Croatian men and women [12], and BMI, WC, and WHtR among Brazilian men and women [13]. Few studies estimated the predictive power of obesity-related indicators by adjusting for the age effect as well. Based on a survey which was conducted in rural areas of northeast China, WHR was the best predictor of new-onset hypertension in both women under 60 and over 60 years, and WHR and WC were the best indices in men aged ≤ 60 and > 60 years, respectively [14]. In another study, it was observed that the predictive power of BMI, WC, WHR, and WHtR was similar in Iranian men in two age groups (20–39 and 40–59 years); however, in Iranian women between 40 and 59 years, WC and WHtR had significantly better predictive power compared to BMI [15]. Based on the literature, in the studies that have been done so far to determine the predictive power of anthropometric indicators based on the area under the receiver operating characteristic (ROC) curve (AUC), the effect of age has been adjusted by categorizing it, and the analysis was performed in each age group. Although the obtained results are valuable, they have several disadvantages. For instance, since the analyses were conducted separately, formal inferences about the effect of age as a covariate are not possible. Furthermore, selecting the cut-off points for the categorization of this continuous covariate is another problem [16]. These drawbacks have led us to use the conditional time-dependent ROC curve. Using this method, this study aimed to investigate the power of obesity-related indices to predict hypertension (HTN) in an Iranian population, considering the effect of age as a covariate in the ROC framework.

Methods

Study design and population

This community-based prospective cohort was initiated with 10,000 Iranian adults between 35 and 70 years old who were recruited in the context of the Shahedieh cohort study in 2016. Shahedieh study is a part of the PERSIAN (Prospective Epidemiological Research Studies in Iran) multicenter cohort study, which was conducted in three cities of Yazd province (Shahedieh, Zarch, and Ashkezar) [17]. More details about the protocol of the PERSIAN cohort study are accessible elsewhere [18]. After the end of the enrollment, the follow-up phase of the cohort study was done annually for 4 years. During this period, participants received annual phone calls, and they were asked about some chronic diseases, such as HTN [19]. This prospective study was designed to investigate the relation of the incidence of HTN with obesity-related indices, which were recorded at baseline.

Sample size and participants

Of 10,000 adults, 2071 individuals who had a previous diagnosis of HTN at the baseline and 136 participants with incomplete information were excluded from the study.

Anthropometric measurement

Obesity-related indices (Body mass index (BMI), waist circumference (WC), wrist circumference and Hip circumference, Waist- To- Height Ratio (WHtR), and Waist-to-hip Ratio (WHR)) were measured by a trained researcher at the baseline. To measure weight, a digital scale (SECA, model 755, Germany) was used while the subjects wore minimal clothing and without shoes. Hip circumference and waist circumference were also measured while participants wore minimal clothing. The height was measured with a precision of 0.5 cm using a tape measure, which was attached to the wall without any bumps [18].

Outcome measurement

Systolic blood pressure (SBP) ≥ 140 mm Hg, or diastolic blood pressure (DBP) ≥ 90 mm Hg, is defined as hypertensive [20]. During 4 years of follow-up time in the Shahedieh cohort study, participants received annual phone calls, and they were asked about some chronic diseases, such as HTN. New cases were individuals who reported the physician-based diagnosis of HTN and were using antihypertensive medication in the annual follow-up phone call [19]. If people declared their HTN, their medical documents were checked and confirmed by a physician. In case of inconsistency, a third internist conducted the assessment of the documents to reach a final decision. In this study, censored participants are defined as those who have not experienced HTN during the follow-up. Furthermore, the interval between the inclusion time and the first positive assessment of HTN was recorded as survival time.

Covariates

Information on demographic characteristics (e.g., age, sex, marital status, and education level) and behavioral measures (smoking and physical activity) was obtained through standardized questionnaires. In the Shahedieh cohort study, SBP and DBP were also measured in a seated position, after at least a 10-minute rest period. The measurements were repeated for both the right and left arms twice, using a mercury sphygmomanometer. The final values of DBP and SBP were obtained by averaging the second blood pressure (BP) measurement of the right and left arms. Cigarette smoking status was also ascertained by the individual’s self-report, so that current smokers were defined as participants who claimed that they smoked regularly or occasionally.

The physical activity of participants was measured through face-to-face interviews using the International Physical Activity Questionnaire, which was translated into Persian and validated for the Iranian population previously [21]. The results of this Questionnaire were expressed in terms of metabolic equivalents per week (MET-h/week). In this study, levels of physical activity were determined as low (24- 36.5 MET-hours per week), moderate (36.6–44.9 MET-hours per week), and heavy (MET ≥ 45 h per week) [22].

Statistical analysis

Continuous and categorical variables were described as mean ± standard deviation (SD) and percentages, respectively. Student’s t-test, Mann-Whitney U test, and Chi-square test were used to compare the baseline characteristics of samples with and without HTN.

For each individual, we recorded baseline marker value, age, survival time (time to hypertension or censoring), and event status. For any prediction time t (here, 4 years), cases were those who developed hypertension within [0, t] and controls were those event-free at least until t. Using the method of Li et al., a kernel smoother with bandwidth selected by least-squares cross-validation gives more weight to individuals whose age is close to the target age z₀. From these conditional sensitivity/specificity estimates, we computed the time-dependent AUC as a smooth function of age, separately for men and women, using the condTimeROC package in R.

In this study, hazard ratios (HRs) and 95% confidence intervals (CIs) were also estimated as the risk of new-onset hypertension for each baseline characteristic using univariate and multivariate (predictors: age, sex, marital status, education level, current smoker, physical activity, SBP, DBP) Cox regression model. Furthermore, to estimate the power of obesity-related indices to predict the incidence of HTN over different ages, AUC was estimated using a conditional time-dependent ROC curve.

The validation of anthropometric indices for clinical use requires assessment of their prognostic capacity for future outcomes. The time-dependent Receiver Operating Characteristic (ROC) curve is a fundamental method for this purpose in survival analysis, quantifying a biomarker’s capacity to distinguish between individuals who will develop hypertension within a specified prediction time and those who will remain event-free. A significant limitation of the standard ROC curve is that it assumes disease status as time-invariant, making it inappropriate for prospective studies of incident disease. Consequently, a time-dependent ROC framework is better suited for such scenarios. Additionally, the predictive performance of biomarkers is often modulated by confounding variables, such as patient age, highlighting the importance of incorporating influential covariates into the ROC analysis [16, 23]. This study employed the optimal data-driven bandwidth method of Li et al., which utilizes least-squares cross-validation. This method computes two distinct bandwidths: A bandwidth for an estimator that smooths only the covariate, corresponding to an unsmoothed nonparametric estimator, and a multivariate bandwidth for an estimator that smooths both the covariate and the marker, corresponding to a smoothed nonparametric estimator [16, 24]. The predictive power of a biomarker is classified as excellent for AUC values between 0.90 and 1, considerable for AUC values between 0.80 and 0.90, fair for AUC values between 0.70 and 0.80, poor for AUC values between 0.60 and 0.70, and failure for AUC values between 0.50 and 0.60 [25].

Results

Of 7793 samples, there were 4133 (53%) males and 3660 (47%) females. The mean age of participants was 46.47 ± 8.86 years at enrollment. During the follow-up period, of 7793 subjects, 303 (3.90%) new cases of HTN were identified. The baseline characteristics, including age, sex, marital status, education level, smoking, SBP, and DBP of the samples, were significantly associated with the risk of new-onset hypertension. The highest incidence of HTN was 9.05% (21/232) in people > 65 years old, and the hazard of hypertension for 45–54 years participants was 1.66 (95% CI 1.23–2.32), for 55–65 years participants was 1.95 (95% CI 1.34–2.84), and for individuals > 65 years was 2.38 (95% CI 1.31–4.33) times the hazard for 35–44 years old group. HTN incidence in women (4.37%) was more than in men (3.45%). The hazard of hypertension for women was 1.63 (95% CI 1.22–2.17) times the hazard for men. In terms of education level, the highest incidence of HTN (6.70%) was observed in illiterates (Table 1).

Table 1.

Baseline demographic and clinical characteristics and their association with the 4-year incidence of hypertension among participants in the Shahedieh Cohort Study, Yazd Province, Iran (baseline enrollment in 2016)

Variables	Total (n = 7793)	Without HTN (n = 7490)	With HTN (n = 303)	P-value	Crude HR (95% CI)	P-value	Adjusted HR * (95% CI)	P-value
Age (years)
35–44	3792 (48.70%)	3716 (49.60%)	76 (25.10%)		Reference		Reference
45–54	2441 (31.30%)	2317 (30.90%)	124 (40.90%)	P < 0.001	2.56 (1.93, 3.41)	P < 0.001	1.69 (1.23, 2.32)	0.001
55–65	1328 (17%)	1246 (16.60%)	82 (27.10%)		3.20 (2.34, 4.38)	P < 0.001	1.95 (1.34, 2.84)	< 0.001
> 65	232 (3%)	211 (2.80%)	21 (6.90%)		4.67 (2.88, 7.57)	P < 0.001	2.38 (1.31, 4.33)	0.004
Sex
men	4133 (53%)	3990 (53.30%)	143 (47.20%)	0.03	Reference		Reference
women	3660 (47%)	3500 (46.7%)	160 (52.80%)		1.26 (1.01, 1.59)	0.03	1.63 (1.22, 2.17)	< 0.001
Marital Status
Single	31 (0.50%)	30 (0.50%)	1 (0.40%)		Reference		Reference
Married	6597 (96.60%)	6332 (96.70%)	265 (93.60%)	0.007	1.28 (0.18, 9.18)	0.80	1.03 (0.14, 7.40)	0.97
Widowed or divorced	200 (2.90%)	183 (2.80%)	17 (6%)		2.88 (0.38, 21.47)	0.30	1.45 (0.19, 11.06)	0.72
Education Level
Illiterate	895 (11.5%)	835 (11.20%)	60 (19.80%)		Reference		Reference
Primary	2374 (30.50%)	2260 (30.20%)	114 (37.60%)	P < 0.001	0.69 (0.51,0.95)	0.02	1.24 (0.87, 1.78)	0.22
Intermediate	1423 (18.30%)	1375 (18.40%)	48 (15.80%)		0.48 (0.33,0.70)	P < 0.001	1.09 (0.69, 1.71)	0.69
Secondary	1750 (22.50%)	1701 (22.70%)	49 (16.20%)		0.40 (0.27,0.58)	P < 0.001	1.11 (0.71, 1.75)	0.62
Higher	1349 (17.30%)	1317 (17.60%)	32 (10.60%)		0.34 (0.22,0.53)	P < 0.001	0.91 (0.56, 1.48)	0.71
Current Smoker
No	6604 (84.70%)	6331 (84.50%)	273 (90.10%)	0.008	Reference		Reference
Yes	1189 (15.30%)	1159 (15.50%)	30 (9.90%)		0.61 (0.42, 0.90)	0.01	0.72 (0.47, 1.10)	0.13
Physical Activity (METs-hours per week)
24-36.5	1700 (21.80%)	1637 (22%)	63 (20.8%)		Reference		Reference
36.6–44.9	4148 (53.20%)	3984 (53.60%)	164 (54.10%)		1.06 (0.79, 1.41)	0.69	1.15 (0.84, 1.56)	0.37
≥ 45	1889 (24.20%)	1813 (24.40%)	76 (25.10%)		1.06 (0.75, 1.48)	0.73	1.35 (0.95, 1.93)	0.08
SBP (mm Hg)	106.49 ± 14.20	105.87 ± 13.70	121.48 ± 17.35	P < 0.001	1.056 (1.05, 1.06)	P < 0.001	1.04 (1.03, 1.05)	< 0.001
DBP (mm Hg)	66.42 ± 10.05	66.10 ± 9.83	74.20 ± 12.02	P < 0.001	1.069 (1.06, 1.07)	P < 0.001	1.01(1, 1.03)	0.03

Abbreviations: HR Hazard ratio, SBP Systolic blood pressure, DBP Diastolic blood pressure, METs Metabolic Equivalents (METs)

*Estimates are from Cox regression models mutually adjusted for all variables (age, sex, marital status, education level, current smoker, physical activity, SBB, DBP)

Out of 303 HTN new cases, 160 (52.80%) were women. The mean and standard deviation (SD) of the baseline obesity-related indices of participants with HTN and participants without HTN are reported in Table 2 by sex. Based on the findings, men and women with HTN had significantly higher mean age, height, weight, BMI, Hip, WC, WHtR, WHR, and wrist circumference than men and women without HTN.

Table 2.

Baseline age and obesity-related indices according to the 4-year incidence of hypertension among male and female participants in the Shahedieh Cohort Study, Yazd Province, Iran (baseline enrollment in 2016)

Characteristics	Men (n = 4133)		P-value	Women (n = 3660)		P-value
	Without HTN (Mean ± SD)	With HTN (Mean ± SD)		Without HTN (Mean ± SD)	With HTN (Mean ± SD)
N (%)	3990 (96.50%)	143 (3.50%)		3500 (95.60%)	160 (4.40%)
Age (year)	47.29 ± 9.07	52.73 ± 8.99	P < 0.001	45.13 ± 8.35	49.93 ± 8.49	P < 0.001
Weight (kg)	78.32 ± 13.84	82.30 ± 14.49	P = 0.001	71.50 ± 12.98	74.69 ± 12.74	P = 0.002
Height (cm)	170.90 ± 6.69	169.40 ± 6.78	P = 0.009	156.79 ± 5.89	154.95 ± 6.01	P < 0.001
BMI (kg/m2)	26.77 ± 4.23	28.66 ± 4.74	P < 0.001	29.07 ± 4.94	31.07 ± 4.74	P < 0.001
Hip circumference (cm)	99.89 ± 7.53	102.11 ± 8.52	P = 0.001	105.09 ± 9.65	107.83 ± 9.89	P = 0.001
WC (cm)	93.50 ± 10.97	98.55 ± 11.75	P < 0.001	95.73 ± 11.68	100.11 ± 10.30	P < 0.001
WHtR	0.54 ± 0.06	0.58 ± 0.07	P < 0.001	0.61 ± 0.07	0.64 ± 0.06	P < 0.001
WHR	0.93 ± 0.06	0.96 ± 0.05	P < 0.001	0.91 ± 0.07	0.93 ± 0.06	P = 0.001
Wrist Circumference (cm)	18.09 ± 1.24	18.42 ± 1.29	P = 0.002	16.47 ± 1.25	16.85 ± 1.28	P < 0.001

Abbreviations: SD Standard Deviation, HTN Hypertension, BMI Body mass index, WC Waist circumference, WHtR Waist- To- Height Ratio, WHR Waist-to-hip Ratio

To assess the power of the studied obesity-related indices to predict HTN, AUC was estimated adjusting for age in both sex groups. Fig. 1 shows the estimated AUC of BMI by gender at different ages, revealing that the discriminatory power of this biomarker was not uniform across all age groups.

Fig. 1.

Age-adjusted time-dependent area under the receiver operating characteristic curve for baseline body mass index as a predictor of 4-year incident hypertension among participants in the Shahedieh Cohort Study, Yazd Province, Iran (baseline enrollment in 2016). Abbreviation: AUC, area under the receiver operating characteristic curve

Figure 1(a) shows the upward trend of AUC for men. Based on this figure, BMI had fair predictive power for men aged 60 to 70 years. The AUC values were less than 0.7 for men under 60. Based on Fig. 1(b), it was also observed that the BMI has fair to excellent predictive power for 36 to 44-year-old and 55 to 60-year-old women. This biomarker had fair to considerable discriminatory power for 47 to 50-year-old women. The AUC values were less than 0.7 for women in other age groups. Figure 2 shows the AUC estimates for Hip circumference for men and women of different ages. Based on the values of AUC in Fig. 2(a), it was observed that Hip circumference had fair predictive power for men around 60. This index had fair to excellent power for HTN prediction in women between 38 and 44 years. The discriminatory power also varied between fair to considerable for 56 to 60-year-old women (Fig. 2(b)). Based on Fig. 3(a), it was observed that WC cannot be used to predict hypertension for men, while this index had fair to excellent predictive power for women in most ages (Fig. 3(b)). Comparing the estimated AUC values with the reference line in Fig. 4 indicates that, unlike WHR, which had an unreliable prediction for men in all ages (Fig. 4(a)), this index had fair to excellent power to predict HTN for women in most of the age groups (Fig. 4(b)). Figure 5 shows the AUC estimates for WHtR for men and women of different ages. The HTN predictive power of this index was negligible (AUC < 0.6) in men. A downward trend in AUC values in Fig. 5(b) also showed that WHtR had fair to excellent prediction power for women less than 58 years old. Figure 6 shows the AUC estimates for wrist circumference at different ages in both sexes. The index showed no predictive value for HTN in men (AUC < 0.6) (Fig. 6(a)). Besides, in women, this index achieved acceptable to outstanding performance in the 47–52 age group and fair to excellent performance in the 57–59 age group (Fig. 6(b)).

Fig. 2.

Age-adjusted time-dependent area under the receiver operating characteristic curve for baseline Hip circumference as a predictor of 4-year incident hypertension among participants in the Shahedieh Cohort Study, Yazd Province, Iran (baseline enrollment in 2016). Abbreviation: AUC, area under the receiver operating characteristic curve

Fig. 3.

Age-adjusted time-dependent area under the receiver operating characteristic curve for baseline waist circumference as a predictor of 4-year incident hypertension among participants in the Shahedieh Cohort Study, Yazd Province, Iran (baseline enrollment in 2016). Abbreviation: AUC, area under the receiver operating characteristic curve

Fig. 4.

Age-adjusted time-dependent area under the receiver operating characteristic curve for baseline Waist-to-hip Ratio as a predictor of 4-year incident hypertension among participants in the Shahedieh Cohort Study, Yazd Province, Iran (baseline enrollment in 2016). Abbreviation: AUC, area under the receiver operating characteristic curve

Fig. 5.

Age-adjusted time-dependent area under the receiver operating characteristic curve for baseline Waist- To- Height Ratio as a predictor of 4-year incident hypertension among participants in the Shahedieh Cohort Study, Yazd Province, Iran (baseline enrollment in 2016). Abbreviation: AUC, area under the receiver operating characteristic curve

Fig. 6.

Age-adjusted time-dependent area under the receiver operating characteristic curve for baseline wrist circumference as a predictor of 4-year incident hypertension among participants in the Shahedieh Cohort Study, Yazd Province, Iran (baseline enrollment in 2016). Abbreviation: AUC, area under the receiver operating characteristic curve

Discussion

Obesity and HTN, which are two major risk factors for CVD, have turned into a threat to public health and contribute to economic burdens. Since the incidence of HTN is increasing throughout the world dramatically with the significant rise in obesity, obesity-related indicators can be used as primary screening tools to prevent future HTN and CVD risk [7]. Over the past few decades, the association between different anthropometric measures and HTN has been studied to investigate the usefulness of anthropometric indices as predictors of HTN in many ethnic groups and countries, primarily based on cross-sectional studies. Since genetic factors may affect this association, detecting appropriate indices to predict HTN differs in each community [15].

In various studies, ROC analysis was used to assess the potential of obesity-related indices to predict HTN and detect the best predictor in each nation [7, 14, 15, 26–28]. The common idea of most of these studies was investigating the overall predictive power of obesity indices without adjusting for the age effect [7, 26–28], while age is simultaneously associated with obesity and hypertension besides sex [7, 9, 10]. To the best of our knowledge, few studies have performed a separate ROC analysis on subgroups defined by age as a potential covariate to adjust the impact of age on the predictive capacity of obesity-related markers [14, 15]. Using the standard ROC analysis is applicable whenever the covariates of interest are categorical or are categorized if they are of a continuous nature. Although this method can certainly provide valuable results, it has several drawbacks. Drawing formal inferences about the effect of age as a covariate is not possible because the analyses are performed separately. Furthermore, there is the problem of selecting the cut-off points to categorise this continuous covariate. These disadvantages have led to the use of conditional time-dependent ROC curve analysis, which includes the effect of age as a covariate in the ROC framework in this study [16]. By estimating AUC by adapting conditional time-dependent ROC curve analysis, the accuracy of each obesity-related indicator to predict HTN was evaluated over age in both sexes in the Iranian population.

Based on the findings, the predictive power of indices varied with age and sex. It was observed that BMI has fair power for predicting HTN in men between 60 and 70, while this index has provided fair to excellent prediction power in most of the women under 60 years old. A study that was conducted on over 60 years of persons in Limpopo province, South Africa, showed that BMI was the best predictor of hypertension in men [29]. Furthermore, the Hip index could predict HTN in men around 60, while this index was practical for predicting HTN in most young and middle-aged women. Our results also indicated that WC could not predict HTN in young and older men, while the predictive power of WC for HTN in women was almost fair to considerable/excellent in most age groups. WC and WHR were detected as better HTN predictors than BMI in women over 60 years in South Africa [29]. According to the findings of our study, it was observed that WHR does not have enough power to predict HTN in men, while the predictive power of this index was fair to excellent for most age groups of women. Our results in this study showed that WHtR and wrist circumference did not provide sufficient accuracy to predict the incidence of HTN in men. The incidence of HTN has fair to excellent prediction in women under 58 based on WHtR values. The predictive power of wrist circumference also fluctuates with age in women. It has fair to excellent prediction power in some age groups of women (47–52 and 57–59). Another study on the middle-aged and elderly Thai population showed that WHtR was the best index for adiposity in association with hypertension [30].

In a study on an Iranian population, the power of obesity indicators to predict HTN in two age ranges of 20 to 39 and 40 to 60 was investigated by gender. According to the findings of this study, the best indicators for HTN prediction in men aged 40 to 60 were WHR and WHtR (AUC = 0.6), and in women in the same age range were WC and WHtR (AUC = 0.64) [15]. In the present study, the range of AUC for WHR in men aged 40 to 60 was 0.37 to 0.54, and it had the best prediction for those aged 47–50 years old with an AUC above 0.5. The range of AUC for WHTR for men in the same age range was 0.42 to 0.59, and the best AUC for this index was observed at the ages of 46–60. In women aged 40 to 60, the range of AUC for WC was 0.56 to 0.99, which is considered acceptable. The range of AUC for WHTR for women in the same age range also showed the same pattern; the values of AUC for this index varied between 0.59 and 0.99.

In another study in China, the predictive power of obesity-related indices was assessed in participants under and over 60 years. Based on the findings of this study, the best indicators for HTN in men < 60 and > 60 years were WC (AUC = 0.75) and WHR (AUC = 0.72), respectively, and in women, in both age ranges, were WHR (AUC = 0.79 and AUC = 0.71, respectively) [14]. In the current study, the range of AUC for WC in men under 60 years was 0.31 to 0.60, and it had the best prediction in 48–60 years, with an AUC above 0.5. The values of AUC for WHR in men over 60 years were less than 0.5 (0.44–0.49). In women under 60 years of age, all AUC values were above 0.5 except at 58 and 59 years old (0.27–0.99). The range of AUC for this index for over 60-year-old women was 0.81 to 0.98. Previous studies showed that, compared with other obesity indices, central obesity indices such as WHR and WC played a more critical role in predicting HTN in the general population. In these studies, the predictive power of obesity-related indices was mainly estimated and reported in terms of a point estimation for AUC for each sex, ignoring age [14, 29, 31], while the findings of our study showed that the accuracy of obesity indicators related to HTN varied with both sex and age. So, our findings in the current study are not comparable with previous studies, except for the ones that were conducted on a specific age group.

The findings of the current study revealed that, for men, only BMI and hip circumference among the obesity-related indices possessed predictive power for HTN, and this was confined to specific age groups. BMI seems clinically applicable for women in specific age bands (36–44 and 55–60 years). However, its performance drops off in other groups, indicating it should be used in conjunction with other measures rather than in isolation. A reason that BMI fails to predict HTN is that this index doesn’t distinguish between fat and muscle mass or account for fat distribution [32].

WHR and wrist circumference were found to be clinically unreliable for men of all ages, with negligible predictive value. WC also showed no utility for HTN prediction. This implies that relying on these simple measures for HTN risk stratification in male patients is not advisable. Clinicians should prioritize other indices for men. Hip Circumference is most relevant for predicting HTN in women of pre-menopausal age (38–44 years) and later middle age (56–60 years). WC emerges as one of the most robust and consistently useful tools for women across most age groups, which reinforces the importance of central adiposity measures in female cardiovascular health. WHtR is highly practical and effective for women who are under 58 years old. Furthermore, for women, wrist circumference serves as a particular tool, showing its best performance in narrow age ranges (47–52 and 57–59 years).

Consistent with previous research, our findings confirm a sex-specific effect in the relationship between obesity measures and hypertension in older adults, reflecting physiological differences in body fat distribution and cardiovascular outcomes between men and women. Future research is needed to determine whether this sex differential stems from biological factors, clinical issues such as physician inertia or patient adherence, or sex-related differences in medication efficacy [33]. In summary, BMI emerged as the most reliable predictor for men, while for women, waist circumference and waist-to-hip ratio performed best across most age groups.

Regarding strengths and limitations, this study is a prospective cohort, while the assessment of the HTN predictive power of different anthropometric measures which were assessed in previous studies was primarily based on cross-sectional studies. The validity of this study was enhanced due to the substantial sample size chosen from a clearly defined population. Shahedieh was conducted on the 35–70-year-old individuals of three small towns; therefore, extending the results to the population aged over 70 years or under 35 years, or to an urban population, needs discretion. However, the ROC curve analysis we used to compare the predictive power of the obesity indices is methodologically limited in its ability to adjust for multiple variables. It can only directly adjust for one continuous covariate. Given this constraint, the age effect was adjusted because it is one of the strongest and most consistent risk factors for hypertension and is also correlated with the obesity indices. Adjusting for other categorical variables (beyond sex, which was included) would require stratifying the sample. With only 303 incident hypertension cases, further stratification would lead to small, unstable subgroups and unreliable AUC estimates. The limited number of HTN cases may have contributed to the lack of smoothness in the estimated age-specific AUC curves, whereas biologically plausible sex-specific differences may explain the observed variation in predictive performance between men and women. In the AUC plot for women. Furthermore, in women, the transition through perimenopause and menopause induces abrupt changes in body fat distribution [34], which may create many spikes and bottoms in the results for women. For instance, the sharp peaks in AUC for waist circumference and WHR around ages 48–52 years (Figs. 3b and 4b) coincide with the average age of menopause in this population. Given the age-specific fluctuations observed for several obesity-related indices, particularly in women, we have to interpret the predictive power of these indices in different age windows with caution.

Furthermore, the ascertainment of incident hypertension was based on annual telephone follow-up and verification of medical records, rather than direct and standardized blood pressure measurements at fixed intervals. It is possible that some participants with undiagnosed, asymptomatic hypertension were incorrectly classified as censored, which could lead to an underestimation of the true incidence. Second, the precise date of hypertension onset was not available. However, the event is known to have occurred within the interval between two follow-up contacts.

Conclusions

In summary, the results investigated the usefulness of obesity-related indices as predictors of HTN in the short term in different sex and age groups, which may have a vital public health implication, especially for the Iranian population. Our results highlight that the predictive power of these indicators depends on both sex and age; however, the AUCs decreased in the elderly, indicating that the ability to recognize HTN gradually weakened with age. This correlation can be due to the presence of confounding variables, such as arterial stiffness, which have a direct relationship with age. Considering the sex and age of individuals, these indices can be utilized as primary screening tools to control obesity, which is one of the main risk factors for HTN, at resource-limited healthcare centers.

Abbreviations

HTN

Hypertension

BMI

Body mass index

WC

Wrist circumference

WHtR

Waist- To- Height Ratio

WHR

Waist-to-hip Ratio

WHO

World Health Organization

MI

Myocardial infarction

CVDs

Cardiovascular diseases

ROC

Receiver operating characteristic

AUC

Area under receiver operating characteristic curve

SBP

Systolic blood pressure

DBP

Diastolic blood pressure

CI

Confidence interval

SD

Standard deviation

Authors’ contributions

R.S., M.H.K.K.L and S.J. conceptualization, methodology, data analysis, write, review and edit; S.M.N.: methodology, clinical analysis, review and edit; S.M.T.: data analysis, reviewing and editing.

Funding

The authors declare that no funds, grants, or other support were received during the preparation of this manuscript.

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

This study was approved by the Ethics Committee of Shahid Sadoughi University of Medical Sciences in Yazd with the code No: IR.SSU.SPH.REC.1401.141. All individuals signed informed consent in accordance with the ethical principles of the Declaration of Helsinki before being enrolled in this study. The authors used AI for editing and grammar improvement. The authors reviewed and verified all AI-assisted content and remain fully responsible for the accuracy, originality, integrity, and ethical compliance of the manuscript.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.