Sex-Specific Cardiometabolic Phenotypes of Metabolic Syndrome Identified by Latent Class Analysis in Indian Adults

Abstract

Background:

Metabolic syndrome (MetS) encompasses a constellation of cardiometabolic risk factors, yet its phenotypic diversity remains underexplored.

Aim:

The study aimed to identify sex-specific latent phenotypes of MetS using key clinical indicators and to evaluate their association with metabolic risk.

Materials and Methods:

We conducted a latent class analysis among 440 adults aged 20–55 years using five indicators – waist circumference, systolic and diastolic blood pressure (BP), high-density lipoprotein (HDL) cholesterol, and Homeostatic Model Assessment of Insulin Resistance (HOMA-IR). Sex-specific models and sensitivity analyses were used to assess measurement invariance and model robustness.

Results:

A three-class model demonstrated optimal fit (lowest Bayesian Information Criterion, entropy ≥0.99). Class 1 (43%) was predominantly low-risk, with minimal abdominal obesity and no hypertension. Class 2 (23%) represented a hypertensive–adiposity phenotype, characterized by universal elevated waist circumference and systolic hypertension but moderate insulin resistance. Class 3 (35%) was marked by pronounced insulin resistance and dyslipidemia, with high HOMA-IR and triglycerides (TG) despite minimal hypertension. Sex-stratified models fit significantly better than pooled models (G² = 130.4, P < 0.001), revealing differential indicator profiles – men showed higher low-HDL prevalence, whereas women exhibited greater diastolic load. Sensitivity analysis collapsing BP indicators confirmed class stability and preserved metabolic gradients. Bolck-Croon-Hagenaars - weighted analysis revealed a stepwise increase in fasting blood sugar and TG across classes.

Conclusion:

Our findings highlight three robust cardiometabolic phenotypes with distinct sex-specific profiles. The hypertensive–adiposity and insulin-resistant classes suggest divergent risk pathways requiring tailored interventions. Even the low-risk group exhibited intermediate lipid elevation, supporting the need for universal lifestyle counseling alongside precision screening strategies.

Introduction

Metabolic syndrome (MetS) is a cluster of interrelated risk factors – including central obesity, insulin resistance, dyslipidemia, and hypertension – that heighten the risk of type 2 diabetes and cardiovascular disease.[1,2] Traditional definitions may overlook heterogeneity within MetS. Latent Class Analysis (LCA), a data-driven approach, enables identification of hidden phenotypic subgroups based on shared metabolic traits without relying on fixed cutoffs.[3,4] This study applies LCA to identify distinct MetS subgroups and examine their associations with insulin resistance, central obesity, and dyslipidemia, with emphasis on sex-specific patterns.[5,6,7,8] Findings aim to inform more targeted, risk-based prevention and management strategies.

Materials and Methods

Study design and population

This cross-sectional study included 440 adults aged 20–55 years, recruited from the medicine outpatient clinics during the period of February 2024 to December 2024. Approval for the study was granted by the Institutional Ethics Committee for Human Research as per the Association Declaration of Helsinki on Ethical Principles for Medical Research.[9]

Inclusion criteria

Eligible participants had at least one component of MetS as per National Cholesterol Education Program Expert Panel (NCEP) and Adult Treatment Panel (ATP) - III criteria and no prior diagnosis of type 2 diabetes or cardiovascular disease.[3]

Exclusion criteria included pregnancy or breastfeeding, history of cardiovascular diseases, cancer, or endocrine disorders (except Type 2 diabetes mellitus [T2DM]), use of medications affecting metabolism (e.g. corticosteroids), incomplete MetS data, or refusal to consent.

Data collection

Data on participants’ sociodemographic profiles, clinical and family histories, and medication adherence were collected from those who provided informed consent. Variables were categorized into six domains: demographics (age, sex, and socioeconomic status), anthropometry (body mass index, waist and hip circumference), clinical (blood pressure [BP], medical history), lifestyle (physical activity, smoking, alcohol intake, and diet via optional questionnaire), and family history (type 2 diabetes, hypertension, and dyslipidemia). Blood samples were drawn after overnight fasting for biochemical analysis, including glucose, insulin, and lipid profile. Derived indices such as Homeostatic Model Assessment of Insulin Resistance (HOMA-IR) were calculated using the following formula: HOMA-IR = (FBG in mg/dL × Fasting Insulin in mIU/L) ÷ 405.[10] Categorical variables were created based on clinical thresholds for LCA. LCA was then conducted to identify subgroups, followed by sensitivity and sex-stratified analyses to test model robustness and measurement invariance. Finally, class-specific metabolic differences were evaluated using the Bolck-Croon-Hagenaars (BCH) method.[11]

Statistical analysis

Latent class indicators and coding

LCA included five metabolic indicators: waist circumference, systolic and diastolic BP (SBP and DBP), high-density lipoprotein (HDL) cholesterol, triglycerides (TG), fasting glucose, and HOMA-IR. Clinical thresholds were applied to derive categorical variables. Waist circumference and BP were coded based on ATP III cut-offs, HDL was sex-specific, and HOMA-IR was categorized into tertiles (T1 ≤ 1.75; T2 = 1.75–3.10; T3 ≥ 3.10).[12] All participants had complete data; no imputation was required.

Latent class modeling

LCA was conducted using the poLCA package in R (v4.3), testing models with 1–6 classes based on the five indicators. Estimation involved 1000 random starts and up to 5000 EM iterations. Model fit was assessed using log-likelihood, Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), and entropy.[5,13] A three-class model was selected based on the lowest BIC and high entropy (median posterior probability = 1.00; Q1 = 0.986), indicating excellent class separation.

Sensitivity analysis

To address potential dependence between SBP and DBP, a binary hypertension indicator (BP) was created (positive if SBP ≥130 mmHg or DBP ≥85 mmHg). The LCA was re-estimated with four indicators (WC, BP, HDL, and HOMA-IR), confirming the robustness of the three-class structure.

Measurement invariance by sex

Separate LCA models for men and women were estimated to assess measurement invariance. A likelihood ratio test indicated significant non-invariance (G² = 130.37, df = 20, P < 0.0001). BIC also favored sex-stratified models (ΔBIC = 37.2), justifying separate analyses for males and females.

Local dependence assessment

Although poLCA does not compute residual χ² statistics, the association between SBP and DBP was evaluated. A strong correlation (Pearson’s χ² = 124.08, P < 0.001) suggested physiological dependence. Despite this, high entropy supported acceptable model performance, with residual dependence acknowledged as a limitation.

Distal-outcome analysis

Class-specific differences in fasting glucose and TG were analyzed using the BCH method. Posterior class probabilities were converted into BCH weights. Using the survey package in R statistical software (version 4.2; R Foundation for Statistical Computing, Vienna, Austria), weighted means and 95% confidence intervals were calculated, and pairwise Wald z-tests were conducted. A Bonferroni-adjusted significance level (α = 0.0167) accounted for multiple comparisons, ensuring valid inference while preserving class separation.

This study applied a robust latent class modeling framework to identify distinct metabolic subgroups using clinically relevant indicators. The use of sex-specific models, sensitivity analysis, and the BCH method for distal-outcome evaluation ensured methodological rigor, minimized bias, and supported the validity and interpretability of the identified classes.

Results

Sample characteristics

Out of 440 participants, 55% were female. The mean age was 37.6 ± 8.2 years. Females had significantly higher HDL levels, whereas males exhibited higher HOMA-IR and BP values.

Latent-class enumeration and profiles

Six unconditional latentclass models (one through six classes) were estimated on the complete‐case sample (n = 440). Table 1 summarizes the global fit indices for all latent class models evaluated. BIC decreased sharply from the 1-to the 3class model, reaching its minimum at three classes (BIC = 2971.6). Adding a fourth class produced only a negligible gain in loglikelihood (Δ =12.0) but increased BIC, AIC, and CAIC, indicating over–fitting. Entropy, a measure of classification certainty was highest for the three-class solution (0.99), confirming excellent separation of latent groups. Based on parsimony, information criteria, and interpretability, the three-class model was retained for all subsequent analyses. Modal assignment yielded the class-size distribution in Table 2.

Table 1.

Fit statistics and entropy for 1–6 latent-class models

Model (k)	LL	ΔLL versus k–1	AIC	BIC	Entropy
1 class	−1566.6	-	3145	3170	-
2 classes	−1472.7	+93.9	2971	3025	0.97
3 classes	−1424.9	+47.8	2890	2971	0.99
4 classes	−1413.0	+12.0	2880	2990	0.95
5 classes	−1406.7	+6.3	2881	3020	0.94
6 classes	−1404.8	+1.9	2892	3059	0.92

Bold values denote the best-fitting model, selected based on minimum BIC/AIC and maximum entropy. AIC: Akaike information criterion; BIC: Bayesian information criterion; LL: Log-likelihood

Table 2.

Complete response-category probabilities for each indicator in the final three-class model (n=440)

Indicator	Response category	Class 1	Class 2	Class 3
WC	Normal (1)	0.000	1.000	0.014
	High (2)	1.000	0.000	0.986
SP (mmHg)	<130 (1)	0.903	0.581	0.000
	≥130 (2)	0.097	0.419	1.000
DP (mmHg)	<85 (1)	0.982	0.838	0.431
	≥85 (2)	0.018	0.162	0.569
HDL	Normal (1)	0.487	0.751	0.621
	Low (2)	0.513	0.249	0.379
HOMA-IR	Tertile 1 (1)	0.334	0.657	0.121
	Tertile 2 (2)	0.371	0.343	0.282
	Tertile 3 (3)	0.294	0.000	0.597

Percentages are conditional on latent-class membership and therefore sum to 100% across categories within each indicator × class block. HOMA-IR: Homeostatic model assessment of insulin resistance; HDL: High-density lipoprotein; SP: Systolic pressure; DP: Diastolic pressure; WC: Waist circumference

The final three-class model assigned all 440 participants with high certainty (median posterior = 1.00; interquartile range = 0.99–1.00; mean = 0.956). Most individuals showed extremely high classification probability (74.1% ≥0.99; 87.3% ≥0.90), reflecting strong class separation. Class sizes and complete response category probabilities are presented in Table 2.

Class 1 (Central Obesity with Normotension) – virtually all members exceeded the waist circumference cut point (100%); systolic elevation was rare (9.7%), diastolic elevation minimal (1.8%); low HDL was common (51%); about one third (29%) were in the highest HOMA-IR tertile, indicating moderate insulin resistance despite normotension.

Class 2 (Emerging Hypertension, Normal Waist) – waist circumference was entirely normal (0% high); 42% showed elevated systolic pressure and 16% elevated diastolic pressure; no member fell into the highest insulin resistance tertile. This pattern suggests early hemodynamic strain prior to central adiposity or marked metabolic dysfunction.

Class 3 (Insulin Resistant Hypertension) – severe hemodynamic burden (100% systolic, 57% diastolic elevation), nearly universal central obesity (99%), and a pronounced biochemical profile with 60% in the highest HOMA IR tertile.

Measurement invariance by gender

To test whether the latent structure was stable across biological sex, the three-class model was re-estimated separately for males (n = 259) and females (n = 181) using identical starting values. Both submodels converged without boundary issues. The summed BIC for the sex specific fits (male BIC = 1773.5; female BIC = 1160.9; total = 2934.4) was 37.2 points lower than the pooled BIC (2971.6), and a likelihood ratio test yielded G² (20) = 130.4, P < 0.001. These statistics indicate modest, but statistically significant, differences in item response probabilities between men and women. Importantly, class enumeration and qualitative interpretation remained unchanged. Each sex specific model reproduced the Relatively Healthy, Hypertensive-Adiposity, and Insulin Resistant Dyslipidemia patterns, with only minor shifts in class prevalence. Accordingly, we retained the pooled model for parsimony, therefore conclude that measurement non-invariance exists, and report sex-specific indicator probabilities in Table 3.

Table 3.

Model-estimated probability of being in the high-risk category for each indicator, stratified by sex and class

Indicator	Class	Men P (high)	Women P (high)
WC	1	0.02	0.02
	2	0.98	0.98
	3	0.03	–
SBP	1	0.00	0.00
	2	1.00	1.00
	3	0.01	0.02
DBP	1	0.48	0.31
	2	0.53	0.66
	3	0.43	0.45
HDL	1	0.78	0.43
	2	0.22	0.57
	3	0.25	0.45
HOMA-IR	1	0.09	0.14
	2	0.28	0.30
	3	0.63	0.56

Note. “P (high)”=model-estimated probability of the high-risk category (row 2 for binaries, tertile 3 for HOMA-IR). Dashes (−) indicate zero probability. HOMA-IR: Homeostatic model assessment of insulin resistance; HDL: High-density lipoprotein; SBP: Systolic blood pressure; DBP: Diastolic blood pressure; WC: Waist circumference

Distal outcomes: Fasting blood sugar and triglycerides

Posterior-rescaled BCH estimates revealed a graded increase in fasting blood sugar (FBS) and TG across classes. Weighted means (±95% confidence interval [CI]) indicated moderate levels in Class 1 (FBS ≈ 134 mg/dL; TG ≈ 146 mg/dL), slightly lower glycaemia and lipids in Class 2 (FBS ≈ 112 mg/dL; TG ≈ 123 mg/dL), and a markedly adverse profile in Class 3 (FBS ≈ 172 mg/dL; TG ≈ 225 mg/dL). Figure 1 summarizes these distributions, while pairwise Wald tests confirmed significant differences between all classes [Table 4; Bonferroni-adjusted P < 0.001]. Notably, despite similar HDL deficits, Class 2 maintained lower FBS and TG than Class 1, whereas Class 3 showed the highest metabolic burden, supporting the clinical relevance of the latent phenotypes.

Figure 1.

Box-and-jitter plots of raw fasting blood sugar and triglycerides distributions by class

Table 4.

Pairwise Wald z-tests (Bonferroni-adjusted; α_adj=0.0167) comparing Bolck-Croon-Hagenaars-weighted means between classes

Variable	Comparison	Δ mean	z	P
FBS	1 versus 2	+22.2	3.59	0.001
	1 versus 3	−38.4	−5.44	<0.001
	2 versus 3	−60.6	−8.47	<0.001
TG	1 versus 2	+23.4	2.64	0.025
	1 versus 3	−78.9	−7.63	<0.001
	2 versus 3	−102.0	−8.77	<0.001

Δ Mean=mean difference (mg/dL). TG: Triglycerides; FBS: Fasting blood sugar

Classification certainty

The three-class latent model demonstrated excellent classification precision. A histogram of participants’ maximum posterior probabilities (30 bins) with an overlaid kernel density curve revealed a pronounced right-skewed distribution, indicating that most individuals were assigned to their latent class with very high certainty. Overall, 74.1% of participants had posterior probabilities ≥0.99, and only 12.7% fell below 0.90, with no values <0.50. Model entropy was 0.99, exceeding the recommended 0.80 threshold and confirming negligible ambiguity in class assignment. These results support the reliability of subsequent distal and comparative analyses by minimizing the risk of misclassification bias.

Local dependence and robustness check: Collapsed-blood pressure model

Because SBP and DBP are physiologically correlated, we evaluated whether residual dependence between SP and DP biased class formation. The pooled three-class model’s expected SP–DP contingency table was compared with the observed table (440 × 2 × 2). A marginal Pearson statistic (Bivariate residual test) gave χ² (1) =124.1, P < 0.001, indicating significant residual association. We recoded a single binary BP (1 = both SBP < 130 and DBP < 85; 2 = either criterion exceeded) and re-estimated the model with four manifest variables (WC, BP, HDL, and HOMA-IR). The best three–class solution reproduced the same qualitative profiles and class prevalences (n = 179/101/160; entropy = 0.96) with a modestly higher BIC (ΔBIC = +6.1) relative to the original fiveindicator model, and Class sizes shifted slightly (55%, 22%, and 23%). Although SP and DP are locally dependent, collapsing them or modeling them jointly did not alter class enumeration or interpretation, suggesting that the latent structure is driven by broader cardiometabolic patterns rather than by any single blood pressure indicator. Nevertheless, potential unmeasured dependencies among other indicators remain a limitation.

A robustness check was conducted by comparing the original five-indicator LCA model (WC, SBP, DBP, HDL, and HOMA-IR) with a collapsed four-indicator model, in which SBP and DBP were combined. Although the collapsed-BP model yielded a lower BIC (2599.0 vs. 2971.6), it showed a slight reduction in entropy (0.97 vs. 0.99) and produced less interpretable class structure. The five-indicator solution retained more balanced and clinically coherent class allocations (42.8%, 22.7%, and 34.5%) compared with the collapsed version (55.0%, 22.0%, and 23.0%), supporting its selection as the final model.

Information criteria are necessary but not sufficient for latent class selection. The collapsed BP model’s BIC advantage (−373 points) was achieved at the cost of markedly lower entropy and elimination of clinically distinct bloodpressure patterns. Consistent with LCA best practice, we adopted the solution that optimized both statistical fit and substantive interpretability. Moreover, clinical differentiation between isolated systolic versus combined systolic–diastolic hypertension, epidemiologically important, was lost. We, therefore, prioritized the five-indicator model, which preserved interpretability and maximized certainty, despite its higher BIC.

Discussion

This study identified three distinct latent cardiometabolic phenotypes using clinically relevant indicators – waist circumference, SBP and DBP, HDL cholesterol, and HOMA-IR – with strong statistical support and high clinical coherence. The model selection was empirically justified through the lowest BIC, high entropy (≈1.0), and sharply delineated posterior probabilities (≥0.99 for 90% of participants). These findings align with prior work demonstrating the utility of LCA in parsing heterogeneous metabolic risk into meaningful subgroups (Gurka et al., 2014).[5,14,15]

The three identified classes reflect divergent yet clinically recognizable pathways of cardiometabolic risk. Class 1, representing a largely “Central - Obesity with Normotension” phenotype, combines near - ubiquitous abdominal adiposity with largely normal hemodynamics. The simultaneous prevalence of low HDL (51%) and moderate insulin resistance (29% in HOMA-IR tertile 3) echoes the “isolated dyslipidemia” phenotype reported in South-Asian cohorts, where visceral fat accumulation precedes overt hypertension and MetS;[16] thus, Class 1 likely represents an early, modifiable stage amenable to waist-centric lifestyle therapy.[17] Although TG were not overtly elevated, intermediate values (mean = 146 mg/dL) suggest subclinical metabolic vulnerability. Prior studies have highlighted the presence of metabolically healthy individuals with isolated lipid abnormalities, cautioning against underestimating risk based solely on obesity or BP.[18,19]

Class 2 characterized by normowaist but moderate-systolic elevation (42%) and negligible insulin resistance resembles the “non-obese hypertensive” profile frequently attributed to salt sensitivity and arterial stiffness rather than adiposity. Its 23% prevalence in our cohort underscores that hypertension in Indian adults cannot be viewed solely through an obesity lens.[20,21] Primary‐prevention strategies for this group may prioritize sodium restriction and vascular‐tone modulation over weight loss. This group illustrates a phenotype where vascular stress precedes glycemic dysregulation. Such clustering of visceral adiposity and BP has been well documented as an early metabolic transition state, often responsive to lifestyle or pharmacologic intervention before the onset of full-blown diabetes (Moore et al., 2017).[22,23] The class’s relatively low HOMA-IR and TG support the hypothesis that early vascular derangements may be independently modifiable risk factors.

Class 3 was characterized by the convergence of central obesity (99%), the highest insulin resistance and triglyceride levels (mean TG = 225 mg/dL; P [HOMA-IR] = 0.60), and severe systolic/diastolic elevation (100%/57%). This mirrors the “combined metabolic–hemodynamic” cluster consistently linked to the highest incident T2DM and cardiovascular events.[24,25,26] This phenotype exemplifies the dysglycemic–lipotoxic pathway, likely rooted in hepatic insulin resistance and adipocyte dysfunction (Nolan et al., 2011).[27,28] The disjunction between BP and metabolic load in this class underscores the limitations of conventional MetS criteria that overly emphasize hypertension as a universal risk flag.[29] Individuals in this group may require early insulin-sensitizing or lipid-lowering strategies, even in the absence of elevated BP.

Sex-specific LCA models further refined our understanding. Separate models significantly outperformed the pooled analysis (G² = 130.4, df = 20, P < 0.001; summed BIC < pooled BIC), confirming measurement noninvariance across sexes. The same latent structure was retained, but indicator expression diverged: women had greater diastolic and HDL burden within the hypertensive-adiposity phenotype, whereas men exhibited more low HDL even in the healthy class. This echoes previous reports that men develop insulin resistance and atherogenic dyslipidemia earlier, whereas women accumulate risk through central obesity and reduced HDL levels.[30,31,32] Pooling the sexes without adjustment risks obscuring such distinct risk profiles, and our results reinforce the need for sex-specific risk stratification in preventive strategies.

To address concerns over the strong correlation between systolic and diastolic BP, a collapsed-BP sensitivity model was evaluated. Despite a significant marginal association (χ² = 124.1, P < 0.001), class structure and metabolic gradients remained robust. While the collapsed model yielded a slightly lower BIC (2599 vs. 2972), it reduced clinical granularity, entropy (0.99 → 0.97), and rebalanced class proportions – indicating that it merged physiologically distinct groups. Furthermore, no additional explanatory power was gained, and the five-indicator model aligns more closely with international MetS definitions that treat SBP and DBP independently.[33] Thus, we retained the five-indicator model as our primary solution.

Importantly, BCH-weighted analyses validated the clinical relevance of these latent classes. A graded metabolic gradient was evident: FBS and TG increased from Class 2 (moderate burden) to Class 1 (intermediate) to Class 3 (highest burden), with all pairwise contrasts remaining significant after Bonferroni correction. These findings counter the assumption that the absence of hypertension and abdominal obesity equates to low metabolic risk. Instead, our “healthy” phenotype still carried moderate dyslipidemia, while the hypertensive-adiposity group had relatively preserved glycemic profiles – suggesting BP elevation may be an early marker of vascular strain rather than overt metabolic dysfunction.

Together, these findings outline two major cardiometabolic trajectories: (1) a vascular-centric path (Class 2) where BP and waist circumference rise early, potentially offering an opportunity for preemptive BP management; and (2) a metabolic-centric path (Class 3) where insulin resistance and dyslipidemia dominate, likely requiring insulin-sensitizing or lipid-modifying strategies. The persistence of these patterns across sensitivity models and sex-specific analyses confirms their robustness and clinical utility.

Limitations and strengths

Several limitations merit consideration. First, the cross-sectional design limits causal inference and temporal tracking of risk evolution. Second, HOMA-IR thresholds were derived from sample-specific tertiles due to the absence of universal clinical cutoffs. While this maximized discriminatory power within our cohort, it may reduce generalisability. Third, residual BP dependence between SBP and DBP was not modeled explicitly, though shown to have minimal impact on class structure.

Nonetheless, this study has considerable strengths: (1) complete data on five clinically relevant indicators; (2) a well-fitting three-class LCA model with high entropy and robust posterior separation; (3) sex-specific invariance testing to guard against bias; (4) BCH-weighted distal analyses ensuring unbiased outcome estimation; and (5) internal sensitivity checks confirming model stability under variable transformations. Collectively, these features enhance both statistical rigor and clinical interpretability.

Clinical implications

Our findings offer clear directions for risk stratification in metabolic disease prevention. Individuals in the “hypertensive-adiposity” class may benefit from early antihypertensive intervention before metabolic decompensation, while those in the “insulin-resistant dyslipidemia” class warrant monitoring for type 2 diabetes and cardiovascular complications. Importantly, the intermediate risk observed in the “healthy” class cautions against complacency, supporting the case for universal lifestyle counseling alongside phenotype-specific interventions. Longitudinal follow-up will be essential to determine whether these latent profiles predict incident diabetes or cardiovascular events.

Conclusion

This study identified three distinct and clinically meaningful cardiometabolic phenotypes through LCA of five key indicators – waist circumference, SBP, DBP, HDL cholesterol, and HOMA-IR. Despite high overall entropy and model stability, sex-specific differences in class expression highlight the need for gender-sensitive risk assessment. Two high-risk phenotypes emerged: a hypertensive–adiposity group with early blood-pressure burden and a metabolically distinct insulin-resistant dyslipidemia group. These classes retained robust separation and clinical relevance across sensitivity analyses, including SBP–DBP collapse.

Our findings suggest two diverging but equally important pathways to cardiometabolic disease – one vascular and one metabolic – that may benefit from targeted prevention strategies. In addition, the substantial portion of individuals classified as “nominally healthy” but exhibiting intermediate lipid disturbances supports the need for universal lifestyle guidance, even in the absence of overt risk markers. Future longitudinal follow-up is warranted to evaluate whether these latent subgroups predict differential risk trajectories for diabetes or cardiovascular outcomes.

Ethical statement

Approval for the study was granted by the Institutional Ethics Committee for Human Research as per the Association Declaration of Helsinki on Ethical Principles for Medical Research (Approval No - PUIECHR/PIMSR/00/081734/4917).

Conflicts of interest

There are no conflicts of interest.

Funding Statement

Nil.