A comparison of anthropometric indicators in predicting the incidence of diabetes and intermediate hyperglycaemia in Finland: a 22-year observational study

Abstract

Background

Given the rising global prevalence of obesity and its strong association with type 2 diabetes, this study aims to evaluate the comparative effectiveness of various anthropometric indicators in predicting the incidence of type 2 diabetes and intermediate hyperglycaemia (IH) over a 22-year period. Due to the limitations of body mass index as a predictive marker, the study assesses alternative metrics, including waist-to-hip ratio (WHR), waist-to-height ratio (WHtR), waist circumference (WC), and relative fat mass (RFM), to determine their predictive power for type 2 diabetes and intermediate hyperglycaemia.

Methods

The cohort comprised 1168 adults (aged ≥ 50 years) from Savitaipale, Finland. Data collection included clinical and laboratory assessments and questionnaires at baseline and at the 10-, and 22-year follow-ups. The incidence of type 2 diabetes and IH was assessed via a 2-h oral glucose tolerance test and health care registry data. Receiver operating characteristic (ROC) curves and area under the curve (AUC) metrics were used to evaluate the predictive power of each indicator, stratified by sex.

Results

WHR had the highest predictive accuracy for type 2 diabetes in men (AUC = 0.70), whereas RFM and WHtR were equally predictive in women (AUC = 0.68). For IH, RFM and WHR were most predictive in men, and WHtR in women. Combining multiple indicators improved the sensitivity of type 2 diabetes prediction.

Conclusions

Alternative anthropometric indicators offer comparable predictive value and show potential for individualised type 2 diabetes and IH risk assessment. Sex-specific cutoffs and a multi-indicator approach could be used to improve screening and early intervention strategies, potentially improving public health management of diabetes risk.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12889-025-25071-3.

Introduction

The prevalence of obesity has increased dramatically worldwide in recent decades, and obesity is currently recognised as one of the greatest challenges to global public health [1]. Rising rates of obesity are associated with an increased incidence of several chronic diseases, particularly type 2 diabetes [2, 3], which affects over 400 million adults and contributes significantly to morbidity, mortality, and healthcare costs [4, 5]. Early detection of individuals at risk is paramount, as timely interventions can prevent progression to type 2 diabetes, and therefore mitigate complications, improve quality of life, and reduce healthcare expenditures [6]. Given the strong link between obesity and type 2 diabetes, early identification and prevention of obesity are essential. Additionally, the increasing prevalence of intermediate hyperglycaemia (IH) is a significant concern as it is associated with an increased risk of cardiovascular disease and premature death [7]. IH often precedes the development of type 2 diabetes, making its early identification and management crucial to the prevention of type 2 diabetes [6–8].

Body mass index (BMI) has long been the most common method of assessing obesity and is currently defined as a BMI ≥ 30 kg/m² in Europid populations [1, 9]. While BMI is easy to calculate, reliable and provides a better estimation of body fat than weight alone, it has notable limitations. It is a nonspecific marker of both body mass and fatness and is unable to differentiate between fat, muscle, and bone mass [9, 10]. Originally developed based on white European men in the mid-1800s, BMI may lack external validity, especially across diverse ethnic groups and in today’s more obese populations [11]. The reliance on BMI as the primary or only indicator of obesity has therefore been increasingly questioned, with calls for more research to compare its ability to predict adverse health outcomes, such as type 2 diabetes, against other anthropometric indicators, particularly in prospective studies [12]. Given the stronger link between visceral fat and the pathogenesis of insulin resistance and type 2 diabetes, other anthropometric screening methods, including relative fat mass (RFM) [13], waist circumference (WC), waist-to-hip ratio (WHR) [14], and waist-to-height ratio (WHtR) [15], have been proposed. This aligns with recent international recommendations, which advise confirming excess adiposity using either direct body fat measurement or an additional anthropometric criterion such as waist circumference, waist-to-hip ratio, or waist-to-height ratio, in addition to BMI, using validated and population-appropriate cut-offs [16].

RFM is a newly developed sex-specific anthropometric equation (based on height and waist measurements) for estimating whole-body fat percentage [17]. WHR is the ratio of waist and hip circumferences, and WHtR is obtained by dividing waist circumference by height.

Few studies with a long-term follow-up have previously compared multiple anthropometric indices simultaneously for the prediction of type 2 diabetes [5, 18]. Additionally, the associations of anthropometric indicators with the development of IH have rarely been investigated. This 22-year observational cohort study, on a population-based sample, provided an opportunity to examine the development of type 2 diabetes and IH both cross-sectionally and longitudinally.

This study aimed to examine the associations of the anthropometric indicators RFM, WC, WHR, and WHtR, with the incidence of type 2 diabetes and IH and to compare their performances with that of a more established indicator, BMI. Given the current gaps in knowledge and the pressing need for effective prediction tools, this study aimed to shed light on the comparative efficacy of these various anthropometric indicators in predicting type 2 diabetes and IH, and their sex-specific differences.

Methods

The Savitaipale study was designed to address the risk factors, prevalence, and incidence of abnormal glucose metabolism and other non-communicable diseases. The study design, including the development and use of the questionnaire used for data collection, has previously been described in detail [19]. The study was conducted according to the Declaration of Helsinki and was approved by the Ethics Review Board of the South Karelia Hospital District. The Research Ethics Committee of the University of Helsinki approved the research (HUS/2203/2018).

Population

The study population included all individuals born 1933–1956, residing in the rural municipality of Savitaipale in Eastern Finland in 1996; all those who consented were included. Written informed consent was obtained from all participants. The target population included 1508 participants, of whom 1168 (581 women and 587 men) participated (attendance rate: 77.5%). The baseline survey was conducted from 1996–1999, with follow-ups conducted from 2007–2008 and again from 2018–2019. Each survey included questionnaires and clinical and laboratory assessments. In addition, data from several national health registers as well as local health care data were obtained.

Baseline data were collected from 1151 participants. By the 10-year follow-up, 75 participants had died and 157 did not participate, leaving 919 participants from whom data were obtained. By the final 22-year follow-up survey, an additional 170 participants had died. Of the remaining people, 704 participated (399 women and 305 men) [19]. Individuals who withheld consent for registry data use were excluded from record linkage (N = 151). Record linkage was also completed for those who died during the study period or did not participate in the follow-up examinations, making case-ascertainment of diabetes virtually complete.

Register data

Information for type 2 diabetes diagnosis, based on the International Classification of Diseases Tenth Revision (ICD-10), was collected from the Finnish Institute for Health and Welfare—Care Register for Health Care (patients discharged from inpatient care and specialised outpatient care), the Social Insurance Institution of Finland (medicine purchases, drug cost reimbursements), and the general practice healthcare information system. These data were obtained for all participants with the computerised record linkage via the personal identification number assigned to all residents in Finland.

Anthropometric measurements

Each survey (baseline, 10-year, and 22-year follow-up) included questionnaires and clinical measurements. The participants had a detailed medical history, physical examinations, and laboratory assessments at the baseline and follow-up examinations. Weight, height, and waist and hip circumferences were measured by a study nurse. Weight was measured with an accuracy of 100 g. Height was recorded with socks on, measured to the nearest 0.5 cm with a single calibrated ruler attached to the wall, ensuring uniformity in the measurement process across all individuals. Waist circumference was measured between the lowest rib and the iliac crest without clothing, and hip circumference from the plane of the trochanters with thin clothing. Both were measured with a flexible tape ruler.

BMI was calculated as weight (kg) divided by height (m) squared at baseline. Baseline height was also used in the 10 and 22-year follow-ups to alleviate the effect of upward bias in the calculation of BMI, since in both men and women, height decreased by approximately 2 cm over the 22-year follow-up period (Supplementary Table 1).

Relative fat mass was calculated according to the following formulae: [20]

WHR was calculated as waist circumference (in cm) divided by hip circumference (cm), and WHtR as waist circumference (cm) divided by height (cm). The cutoff points for measurements are presented in Table 1.

Table 1.

Conventional cutoff points for each anthropometric indicator

	Women	Men
BMI	Underweight: < 18.5	Underweight: < 18.5
	Normal weight: 18.5—24.9	Normal weight: 18.5—24.9
	Overweight: ≥ 25	Overweight: ≥ 25
	Obese: ≥ 30	Obese: ≥ 30 [21]
WC	Increased risk: > 80 cm	Increased risk: > 94 cm
	Substantially increased risk: > 88 cm	Substantially increased risk: > 102 cm [22]
RFM	High risk ≥ 40%	High risk ≥ 30% [17]
WHR	Substantially increased risk: ≥ 0.85	Substantially increased risk: ≥ 0.90 [22]
WHtR	Increased risk: 0.5	Increased risk: 0.5 [23]

Abbreviations:BMI Body mass index, WC Waist circumference, RFM Relative fat mass, WHR Waist-to-hip ratio, WHtR Waist-to-height ratio

Diabetes and intermediate hyperglycaemia

The surveys included the standard 2-h oral glucose tolerance test (OGTT) to investigate the prevalence and incidence of type 2 diabetes and IH and their changes. The OGTT with 75 g of anhydrous glucose was performed in persons without history of type 2 diabetes and with fasting plasma glucose (FPG) < 8 mmol/l. The World Health Organisation (WHO) 1999 classification for type 2 diabetes and IH, including FPG, 2-h plasma glucose, impaired fasting glucose and impaired glucose tolerance, was applied [24]. People using glucose-lowering drugs were classified as having diabetes. A fasting venous blood glucose (FBG) specimen was drawn and analysed on the survey site at baseline and 10-year follow-up. The FBG result was converted to plasma glucose using a factor of 1.12. In the 22-year follow-up, plasma glucose results were obtained from the South Karelia Central Hospital laboratory.

Participants were included in the IH analysis if they were free of type 2 diabetes at baseline (based on OGTT and register data), participated in the follow-up OGTT, and had not developed type 2 diabetes prior to the follow-up examination (based on register data). Individuals who developed type 2 diabetes between baseline and a follow-up survey were excluded from the IH analysis, as they were no longer eligible for OGTT-based reassessment. As a result, type 2 diabetes and IH were analysed as mutually exclusive outcomes. IH classification was applied only to those who remained diabetes-free until their follow-up OGTT.

Statistical methods

Data were analysed using IBM SPSS Statistics version 28.0.1.0 and SAS software version 9.4. Statistical significance was set at P < 0.05. Pearson's correlation (r) was used to determine the correlations among the anthropometric indicators.

Receiver operating characteristic (ROC) curves were used to assess the discriminatory power of each anthropometric indicator in predicting type 2 diabetes. ROC analysis was performed using the continuous values of each anthropometric indicator. In the ROC analyses, each individual value of the conventional and standard anthropometric indicators was used as a threshold to calculate the sensitivity and 1-specificity of the outcome. A ROC curve was then constructed based on these points. Area under the curve (AUC) values were calculated to compare the overall performance of the indicators.

For these analyses, a binary outcome framework was applied. Participants were classified as cases if they were found to have type 2 diabetes, determined either by an OGTT at baseline or from register-based diagnoses during the 22-year follow-up. Time-to-event methods were not used in the calculation of AUC, sensitivity or specificity. Participants who were diagnosed with type 2 diabetes during follow-up were treated as non-cases up to the point of diagnosis and were included as cases thereafter.

For descriptive and cumulative incidence analyses, each anthropometric indicator was divided into four sex-specific groups, based on standard deviations (SD) from the mean: ≤ −1 SD, −1 to 0 SD, > 0 to 1 SD, and > 1 SD from the mean. This classification was used for reporting the prevalence of type 2 diabetes at baseline by sex for each anthropometric indicator.

In addition, to compare predictive performance using a binary approach, we applied sex-specific > 1 SD thresholds for each indicator. The specific mean > 1 SD cutoffs were as follows: for women, BMI > 31.7, RFM > 42.8%, WC > 97.9 cm, WHR > 0.92, and WHtR > 0.61; and for men, BMI > 30.4, RFM > 30.6%, WC > 106 cm, WHR > 1.02, and WHtR > 0.60. A score > 1 SD from the mean was used to create a standard scale for comparison among the anthropometric indicators and to decrease outlier bias that may arise when comparing crude values. With this standardised approach, the relative risk associated with each anthropometric indicator could be compared with others directly.

Sensitivity and specificity were calculated using both standardised (> 1 SD) and conventional clinical cutoffs (Tables 2 and 3). Conventional cutoffs (Table 1) followed established clinical thresholds.

Table 2.

Predictive power of anthropometric measurements at baseline for diabetes in women (N = 508)

Anthropometric measurement	Cutoff point	Cases over cutoff point n	Incidence of T2D n (%)	Sensitivity (%)	Specificity (%)
Standardised BMI > 1	31.7 kg/m2	74	42 (56.8)	24.7	90.5
Standardised WC > 1	97.9 cm	72	45 (64.3)	26.8	92.6
Standardised RFM > 1	42.8%	79	48(62.3)	28.6	91.4
Standardised WHR > 1	0.92	60	40 (55.6)	23.8	90.5
Standardised WHtR > 1	0.61	70	43 (59.7)	25.6	91.4
BMI > conventional risk limit	30 kg/m2	79	56 (55.4)	32.9	86.7
WC > conventional risk limit	89 cm	103	89 (49.4)	53.0	72.9
RFM > conventional risk limit	40%	92	82 (49.1)	48.8	74.7
WHR > conventional risk limit	0.85	386	94 (48.0)	56.0	69.6
WHtR > conventional risk limit	0.5	394	135 (43.4)	80.4	47.6

Table 3.

Predictive power of anthropometric measurements at baseline for diabetes in men (N = 504)

Anthropometric measurement	Cutoff point	Cases over cutoff point n	Incidence of T2D n (%)	Sensitivity (%)	Specificity (%)
Standardised BMI > 1	30.4 kg/m2	74	44 (59.5)	23.7	90.6
Standardised WC > 1	106 cm	72	46(63.9)	24.7	91.8
Standardised RFM > 1	30.6%	79	53 (67.1)	28.5	91.8
Standardised WHR > 1	1.02	60	42 (70.0)	22.6	94.3
Standardised WHtR > 1	0.6	70	49 (70.0)	26.3	93.4
BMI > conventional risk limit	30 kg/m2	79	47 (59.5)	25.3	89.9
WC > conventional risk limit	103 cm	103	61 (59.2)	32.8	86.8
RFM > conventional risk limit	30%	92	57 (62.0)	30.6	89.0
WHR > conventional risk limit	0.9	386	164 (42.5)	88.2	30.0
WHtR > conventional risk limit	0.5	394	167 (42.4)	89.8	28.4

Tables 2 and 3 show the ability of anthropometric indicators to predict type 2 diabetes in women and men over the 22-year follow-up period, using the specific > 1 SD values for each indicator. The incidence of type 2 diabetes using conventional cutoffs for these indicators is also shown in Tables 2 and 3

Abbreviations: Type 2 diabetes (T2D), Body mass index (BMI), Waist circumference (WC), Relative fat mass (RFM), Waist-to-hip ratio (WHR), Waist-to-height ratio (WHtR)

The cumulative incidence of type 2 diabetes was determined as follows. First, individuals with known diabetes (i.e. a prior diagnosis of type 2 diabetes) at baseline and those without consent to the registry data linkage were excluded. Participants who received a new diagnosis of type 2 diabetes based on OGTT results during the baseline examination were included in the incidence analyses, as they had no documented history of diabetes prior to the study entry. These individuals were classified as incident cases, with their date of diagnosis defined as the date of the baseline examination. Excluding such cases would introduce selection bias and underestimate early incidence.

The Kaplan–Meier method was subsequently used to calculate the cumulative incidence of type 2 diabetes. The occurrence of new type 2 diabetes in the registry data or in the three OGTTs was considered as incident type 2 diabetes. Deceased people cases with a cause of death other than type 2 diabetes were censored at the date of the death. People with no type 2 diabetes by the end date of the 22-year study follow-up were censored at that date. We assessed the incidence of IH from baseline to the 10-year follow-up for each anthropometric indicator in men and women who were free of type 2 diabetes at baseline.

As a sensitivity analysis, we also calculated the cumulative incidence of type 2 diabetes based exclusively on registry-based diagnoses, excluding all OGTT-based diagnoses at baseline and during follow-up. This approach aimed to capture purely prognostic incidence, distinguishing it from cases identified through baseline OGTT screening. Kaplan–Meier curves were generated and stratified into four sex-specific groups, based on one SD from the sex-specific man for each anthropometric indicator.

Results

Characteristics of the study population

The general characteristics of the study population are presented in Supplementary Table 1. The number of participants decreased over the 22 years. For the “all participants at each survey” group, weight remained relatively stable, while for the cohort of participants who participated in all follow-up surveys, weight increased steadily by 2 kg in both sexes. While the mean BMI remained relatively stable over the 22 years, the mean WC increased numerically in both women and men. WHtR and RFM also increased in both men and women over the 22 years. The increase in anthropometric indicators was more pronounced in the cohort of people who participated in all the surveys than changes in all participants at each survey point. Thus, the body shape of the participants was changed with age.

Of the initial study population, individuals with baseline known type 2 diabetes were excluded, and the analysis was limited to the 991 individuals who consented to the use of their registry information. Of these participants, 173 developed type 2 diabetes during the 10-year follow-up and 347 during the 22-year follow-up, with an overall incidence rate of type 2 diabetes at 35%. A considerable proportion of those who died during the follow-up had diabetes; 32% and 41% during the 10-year and 22-year follow-ups, respectively. The prevalence of type 2 diabetes increased from 8 to 38% during the 22-year period, similarly in men and women. The proportion of normoglycaemic people decreased from 73% at baseline to 44% at the 22-year follow-up.

Prediction of type 2 diabetes

The highest incidence of type 2 diabetes was observed in women with WC > 97.9 cm (64.3%) and RFM > 42.8% (62.3%), and in men with WHR > 1.02 (70.0%) and WHtR > 0.60 (70.0%) (Tables 2, 3 and 4).

Table 4.

AUC and its 95% CI for Receiver Operating Characteristics (ROC) curve in anthropometric measurements as exposures and diabetes at the 22-year survey as the outcome

Anthropometric measurement	Women		Men
	AUC	95% CI	AUC	95% CI
BMI	0.65	0.60–0.70	0.64	0.59–0.70
WC	0.67	0.62–0.72	0.67	0.62–0.72
RFM	0.68	0.63–0.73	0.69	0.64–0.73
WHR	0.67	0.62–0.72	0.70	0.65–0.75
WHtR	0.68	0.63–0.73	0.69	0.64–0.73

Abbreviations: BMI Body mass index, WC Waist circumference, RFM Relative fat mass, WHR Waist-to-hip ratio, WHtR Waist-to-height ratio

In addition to AUC values, sensitivity and specificity were evaluated for each indicator (Tables 2 and 3). Among standardised cutoffs, RFM had the highest sensitivity in both women (28.6%) and men (28.5%), with specificity above 90% for all indicators. For conventional cut-offs, WHtR > 0.5 showed the highest sensitivity (80.4% in women, 89.8% in men), though with lower specificity (47.6% and 28.4%). These results illustrate the trade-offs between cutoff methods and support the use of multiple indicators for more balanced risk detection.

The ROC curves for each anthropometric indicator’s ability to predict type 2 diabetes during the 22-year follow-up are presented in Supplementary Fig. 1, with the AUCs and their 95% CIs listed in Table 4. While the differences in ROC curves and AUCs across the indicators were generally small, significant differences were observed primarily in comparisons involving BMI (Supplementary Table 2). For example, differences between BMI and RFM were significant in both men (p = 0.006) and women (p = 0.029), suggesting that BMI may have limitations compared to other indicators. All AUC values were greater than 0.5, indicating that each indicator had some discriminatory capacity; however, none reached 0.8, meaning obesity alone was not an adequate predictor of type 2 diabetes.

Figure 1 presents Kaplan–Meier curves for individuals with > 1 SD for each anthropometric indicator. In women, those with high RFM, WC, and WHtR developed type 2 diabetes earlier than those with high BMI or WHR. In men, differences among the indicators were minimal.

Fig. 1.

The cumulative incidence of T2D by anthropometric indicators during 22-year follow-up, stratified by sex

The cumulative incidence of type 2 diabetes (T2D) in individuals with values one standard deviation above the mean for each anthropometric indicator at baseline is shown for both men and women. Type 2 diabetes was defined using the data from each of the three survey examinations or the data from the health registers during the follow-up period. The cutoff point for each indicator is provided with parentheses, “n” denotes the number of cases above the cutoff point, and the percentage in parentheses represents the proportional share of cases within these study participants. The duration is provided in years

Correlations among all pairs of anthropometric indicators were high in both sexes (Supplementary Table 3). In women, WHtR and RFM had a perfect linear association. After that, the highest correlation was seen between WC and WHtR (0.97) and the weakest between BMI and WHR (0.60). In men, WHtR and RFM also had a perfect linear association (1.00). Following WHtR and RFM, the strongest correlations were seen between WC and WHtR (0.94) and WC and RFM (0.94). The weakest correlation was found between BMI and WHR (0.63).

The prevalence of previously undiagnosed type 2 diabetes increased significantly over the 22-year period across all anthropometric indicators (Supplementary Table 4). In women, the greatest difference between baseline prevalence and 22-year prevalence was seen in WC, while BMI showed the smallest change. In men, the greatest difference was seen in WHR, whereas the smallest in BMI.

Prediction of intermediate hyperglycaemia and type 2 diabetes

The incidence of IH and type 2 diabetes, assessed across standardised categories of each anthropometric indicator, showed a consistent pattern: higher standardised values were associated with a greater combined incidence rates of IH and type 2 diabetes (Fig. 2). Among men, RFM, WHR, and WHtR showed the highest combined incidence at > 1 SD above the mean. In women, WC had the highest combined incidence, followed closely by BMI, RFM, and WHtR, which showed relatively similar rates.

Fig. 2.

Incidence of intermediate hyperglycaemia and type 2 diabetes at the 10-year follow-up by anthropometric indicators and sex. The figure shows the combined incidence of intermediate hyperglycaemia and type 2 diabetes from baseline to the 10-year follow-up survey for individuals with anthropometric indicator values across standardised categories for each anthropometric indicator at baseline

Combined predictive power

We examined the joint prediction of type 2 diabetes for BMI, RFM, and WC using a Venn diagram (Supplementary Fig. 2). In men, out of 60 total cases, 53 were identified with RFM > 1 SD above the mean, 46 with a WC > 1 SD, and 44 with BMI > 1 SD. RFM alone increased the predictability by nine cases, BMI by two and WC by three. In women, out of 57 total cases, 48 were identified with RFM > 1 SD above the mean, 45 with a WC > 1 SD, and 42 with BMI > 1 SD. BMI alone increased the predictability by six cases, RFM by four and WC by one. RFM performed better in men than women, identifying 88.3% of cases versus 84.2%. Additionally, as shown in Supplementary Table 5, the combination of different anthropometric indicators, such as RFM or WC > 1 SD, improved sensitivity compared to individual indicators, further supporting the advantage of using a multi-indicator approach for type 2 diabetes risk assessment.

Cumulative incidence curves

The cumulative incidence of type 2 diabetes by baseline anthropometric categories is presented in Supplementary Fig. 3. For all indicators, the incidence was consistently highest in the > 1 SD group, with particularly steep increases observed in men after year 10.

In the sensitivity analysis that excluded baseline OGTT-detected cases (Supplementary Fig. 4), the overall patterns remained similar. However, the separation between risk groups became more pronounced, especially for WHtR and RFM in men.

Discussion

Obesity is widely recognised as the leading risk factor for type 2 diabetes, but how best to assess it remains debated. The increasing prevalence of type 2 diabetes worldwide necessitates the identification of reliable and easily accessible markers for early detection and risk stratification. While BMI has traditionally been the go-to metric for obesity assessment, its limitations, especially in distinguishing fat from muscle mass and its applicability over diverse populations, are well-documented. Our study, spanning over two decades, provides compelling evidence that alternative anthropometric indicators may offer additional predictive power. Our findings are consistent with recent studies that have highlighted the potential of other anthropometric indicators, particularly RFM [14] and WHtR [10, 25], in predicting type 2 diabetes. WC also demonstrated comparable predictive performance to BMI, in line with previous research [26]. Large-scale systematic reviews and pooled cohort analyses have further shown that waist-related measures such as WC, WHtR, and WHR are more strongly associated with type 2 diabetes risk than BMI [5, 27]. Our study supports these observations and adds long-term prospective data and sex-specific performance metrics for a wide range of indicators, including the newer measure RFM, in a general European population.

Our results indicated WHR as the most predictive metric for type 2 diabetes in men, while RFM and WHtR were equally predictive for women. Overall, the differences in predictive power across indicators were relatively small among individuals with values > 1 SD above the mean. All AUC values exceeded 0.5, showing some discrimination capacity, though none reached the ≥ 0.8 threshold. The highest AUC was 0.70 for WHR in men and 0.68 for RFM and WHtR in women (Table 4). While the differences between indicators were generally small, BMI tended to perform worse in direct comparisons, with statistically significant differences observed primarily in comparisons involving BMI (Supplementary Table 2). These differences suggest that sex-specific cutoffs could enhance predictive accuracy, as body size and shape vary between sexes. High values across different indicators mostly identified the same individuals, however, our combination analysis showed improved sensitivity for type 2 diabetes prediction when multiple anthropometric indicators were used together, supporting a multi-indicator approach for more effective risk assessment.

In men, RFM, WHR, and WHtR showed the highest combined incidence of IH and type 2 diabetes in the > 1 SD above the mean group, while WC showed the highest incidence in women, followed closely by BMI, RFM, and WHtR. These results highlight the importance of considering sex-specific cutoffs for IH risk prediction as well.

Beyond type 2 diabetes, these findings might have implications for other obesity-linked conditions. If RFM and WHtR better reflect visceral fat, their predictive power might extend to cardiovascular diseases or metabolic syndrome. Their simplicity could also make them particularly valuable in low-resource settings, where advanced tools are inaccessible. Given the rising prevalence of obesity globally, there is a need for a measurement tool that is easy to use, applicable and possesses adequate sensitivity and specificity. Despite BMI’s limitations, it remains widely used [28, 29].

We saw a significant spike in the cumulative incidence of type 2 diabetes at the 10-year follow-up in men (Supplementary Fig. 3), likely reflecting differences in healthcare-seeking behaviour. Men tend to seek medical care less frequently than women, potentially delaying type 2 diabetes diagnosis until symptoms are more apparent or until routine screenings, such as the 10-year OGTT. This argues for targeted screening for type 2 diabetes in obese middle-aged men, for example via health check-ups, as it could facilitate earlier detection and management.

The main strength of our study is its population-based design with a long follow-up period. We achieved complete ascertainment of the primary outcome, type 2 diabetes, and assessed IH using OGTT. Register data accounted for type 2 diabetes in deceased participants and those who missed follow-up surveys, making this is one of the most comprehensive prospective studies to examine the incidence of type 2 diabetes and IH. Results from participants who completed both follow-up examinations were consistent with the overall cohort, mainly because we were able to ascertain incident type 2 diabetes through record linkage, even for those individuals who did not participate in the follow-up examinations. We compared the effects of several anthropometric indicators simultaneously, including the newly proposed ones.

Further research is needed to validate our findings in more diverse populations and age groups, as our sample consisted largely of white Europeans aged ≥ 50 years. On the other hand, this is the age range when glucometabolic problems primarily occur [29]. Additionally, exploring the combination of anthropometric indicators with other metabolic markers might enhance predictive accuracy. While this study was longitudinal, with a long follow-up of 22 years, it was based on a relatively small population sample, and performing similar investigations on larger cohorts would be beneficial.

While obesity is one of the most common causes of type 2 diabetes, other factors like ageing also contribute to its development. In our cohort, WC, WHtR, and RFM increased over time, even as BMI remained relatively stable, indicating that body composition and fat distribution changed with age (Supplementary Table 1). Survival bias can occur in older populations, as individuals with poorer health are more likely to die during follow-up. To account for this, we separately examined those who completed all follow-ups, though this reduced the cohort size (Supplementary Table 1). The overall trends were broadly similar between the two groups, suggesting that survival bias had a limited impact on the observed changes over time.

An additional limitation is that population-based standardised values can only be determined if data on height, weight, waist circumference, and their calculated averages and variances are available for the population. However, once such data are available, population-specific risk thresholds can be used instead of global cutoffs. This is important because body structures can vary greatly across different populations.

While the inclusion of participants diagnosed with type 2 diabetes via baseline OGTT improved case ascertainment by identifying previously undiagnosed individuals and improving the completeness of incidence estimates, it also introduced methodological challenges for time-to-event analysis. Although these individuals were unaware of having type 2 diabetes before study entry, they were not strictly disease-free at baseline and may have modified their behaviour following the OGTT result. However, completely excluding them would have introduced selection bias by omitting untreated, undiagnosed cases. This highlights how the use of OGTTs, while a strength in terms of diagnostic precision, can also complicate the interpretation of long-term predictive analyses.

To further distinguish between diagnostic and prognostic aims, we conducted a sensitivity analysis excluding baseline OGTT-detected cases. This approach more closely reflected a prognostic study design and revealed an even greater visual distinction between risk groups based solely on registry-confirmed type 2 diabetes incidence, independent of baseline OGTT detection.

Conclusion

In conclusion, while BMI remains a valuable tool to predict type 2 diabetes, alternative anthropometric indicators, especially RFM and WHtR, along with a multi-indicator approach for type 2 diabetes, should be considered for the early detection and risk stratification of type 2 diabetes and IH. Sex-specific differences in predictive accuracy of each indicator should be considered when developing targeted screening and intervention strategies. As the global burden of type 2 diabetes continues to rise, using these simple yet effective tools to predict glucose metabolism disorders can significantly help mitigate its impact.

Abbreviations

BMI

Body Mass Index

WHR

Waist-to-Hip Ratio

WHtR

Waist-to-Height Ratio

WC

Waist Circumference

RFM

Relative Fat Mass

IH

Intermediate Hyperglycaemia

ROC

Receiver Operating Characteristic

AUC

Area Under the Curve

T2D

Type 2 Diabetes

OGTT

Oral Glucose Tolerance Test

FPG

Fasting Plasma Glucose

WHO

World Health Organization

ICD-10

International Classification of Diseases Tenth Revision

FBG

Fasting Blood Glucose

SD

Standard Deviation

SPSS

Statistical Package for the Social Sciences

SAS

Statistical Analysis System

Authors’ contributions

J.L., K.S., J.S., and J.T. contributed to the conception and design of the study. J.L. and M.K. performed the data analysis and contributed to the interpretation of results. J.L. drafted the manuscript, and all authors critically reviewed and revised it for intellectual content. All authors have approved the final version of the manuscript and agree to be personally accountable for their own contributions and to address questions related to the accuracy or integrity of any part of the work.

Funding

Open Access funding provided by University of Helsinki (including Helsinki University Central Hospital). This work was supported by the Savitaipale community and health centre, EVO grants, the Wellbeing Services County of South Karelia, Viipurin tuberkuloosisäätiö, and Elsemay Björn Fund.

Data availability

The datasets generated and/or analysed during this study are not publicly available due to participant confidentiality and privacy concerns. However, data can be accessed from the corresponding author upon reasonable and in accordance with the ethics approval granted for this study.

Declarations

Ethics approval and consent to participate

The study was conducted according to the Declaration of Helsinki and was approved by the Ethics Review Board of the South Karelia Hospital District. The Research Ethics Committee of the University of Helsinki approved the research (HUS/2203/2018).

Written informed consent was obtained from all participants prior to their inclusion in the study.

Consent for publication

Not applicable.

Competing interests

J.T. owns stocks in Orion Pharma, Aktivolabs LTD and Digostics LTD. The other authors declare no potential conflict of interest relevant to this article.