Insufficient physical activity and diabetes mellitus prevalence in Iran: an estimated population attributable fraction analysis using STEPS 2021 data

Abstract

Insufficient physical activity (IPA) affects over 50% of Iranian adults and is a modifiable risk factor for diabetes mellitus (DM). This study estimates the population attributable fraction (PAF) of DM associated with domain-specific IPA in Iran. This cross-sectional study analyzed Iran STEPs 2021 data from 18,013 adults aged ≥ 18 years. IPA was defined as < 600 MET-minutes/week in leisure-time or walking domains using GPAQ-2. Age- and sex-specific PAFs were calculated using Miettinen’s formula with external relative risks from meta-analyses. Domain-specific PAFs are non-additive. DM prevalence was higher among adults with leisure-time IPA versus active peers (19.6% vs. 10.5%, p < 0.001) and among those with walking inactivity versus active peers (18.8% vs. 17.9%, p = 0.017). Using external relative risks, 14.6% (95% CI 11.5, 17.6) of DM cases were associated with leisure-time IPA and 15.7% (95% CI 12.4, 18.9) with walking inactivity. The internal age-sex adjusted odds ratio for walking inactivity was 1.11 (95% CI 1.02, 1.21). Women had higher PAFs than men (leisure-time: 15.0% vs. 14.1%; walking: 16.4% vs. 14.6%). Leisure-time IPA prevalence exceeded 95% among adults with no formal education, women, and older adults. Approximately 50% of individuals with DM remained untreated. Under causal assumptions, approximately 15% of DM cases in Iran were associated with IPA. If these associations reflect modifiable causal relationships, increasing physical activity particularly leisure-time activity could lower diabetes prevalence. However, near-universal inactivity in several subgroups suggests activity promotion should be embedded within comprehensive strategies addressing obesity, diet, and healthcare access. These estimates are hypothetical impact fractions, not precise predictions of preventable cases.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1038/s41598-026-57302-x.

Introduction

Insufficient physical activity (IPA) is a well-established and modifiable risk factor for a wide range of non-communicable diseases (NCDs), including diabetes mellitus (DM), ischemic heart disease, and stroke¹. Despite extensive health promotion efforts, the global prevalence of IPA has increased significantly in recent decades. Based on pooled data from 507 population-based studies conducted across 163 countries, approximately 31.3% of adults worldwide were insufficiently physically active in 2022, compared to 23.4% in 2000². This rise has been particularly notable among women and older adults across all regions. According to national reports from the Iran STEPs 2021 survey, Iran has been classified among countries with moderate to high levels of IPA, with a reported prevalence of 51.3% in 2021³.

Among NCDs linked to IPA, DM is a major contributor to the global burden of disease and premature mortality⁴. Recent epidemiological data indicate a significant rise in the global burden of DM over the past few decades, mainly due to a steady increase in its incidence. In 2021, the Middle East and North Africa (MENA) region reported the highest global age-standardized prevalence of DM at 9.3%, projected to rise to 16.8% by 2050⁴. In Iran, the age-standardized prevalence of DM was estimated at 6.8%, with a disability-adjusted life year (DALY) rate of 1.0% in 2019⁵. Ample evidence from both interventional^6–8 and observational^9,10 studies indicates that adequate physical activity can significantly decrease the likelihood of DM onset. However, in low- and middle-income countries such as Iran, the contribution of IPA to the national burden of DM remains inadequately characterized.

Given the substantial burden that IPA places on both public health and healthcare systems¹¹, it is important to adopt reliable measures for estimating its impact on common chronic diseases. One widely used epidemiological approach for this purpose is the population attributable fraction (PAF)¹². This metric helps quantify the proportion of disease burden, such as DM, that can be linked to specific risk factors, including IPA. The PAF combines the relative risk (RR) of a condition with the prevalence of the associated risk factor in the population, offering a quantitative estimate of how much that risk factor contributes to the disease outcome under causal assumptions. Consequently, the PAF serves as a key tool for evaluating the population-level health impact of IPA and for assessing the potential success of interventions intended to enhance physical activity. It is important to note that PAF estimates derived from cross-sectional data rely on strong assumptions, including the transportability of external relative risks and the absence of reverse causation¹².

While much of the existing literature has relied on global estimates, recent analyses from the MENA region demonstrate that DM now constitutes the largest share of the disease burden attributable to low physical activity¹³, highlighting the need for nationally representative PAF estimates to contextualize this burden within Iran. However, previous estimates for the region have largely depended on relative risks derived from non-MENA populations, and few studies have examined domain-specific physical activity (leisure-time versus walking) separately^13,14.

Since 2005, Iran has implemented a national survey system grounded in the World Health Organization’s STEPwise approach to risk factor surveillance (STEPs), aiming to generate consistent and up-to-date data to inform health policy decisions¹⁵. The physical activity patterns of Iranian adults have been previously evaluated and documented using data from both the 2016 and 2021 STEPs surveys^3,16. These surveys provide a unique opportunity to examine domain-specific IPA and its association with objectively measured diabetes prevalence at the national level.

This study primarily aimed to examine the prevalence of domain-specific IPA among Iranian adults, with a focus on its association with demographic and health-related factors, including DM status. A key objective was to estimate the PAF of DM associated with IPA during leisure-time and walking activities, with analyses stratified by age group and sex, while acknowledging the methodological limitations inherent in cross-sectional PAF estimation. The present study addresses this gap by providing the first domain-specific PAF estimates for DM attributable to insufficient leisure-time and walking activity in Iran using nationally representative data, thereby informing evidence-based diabetes prevention policy.

Materials and methods

Study design

This study was conducted based on the WHO STEPwise approach to non-communicable disease risk factor surveillance (STEPs), version 3.2¹⁷. The detailed procedures and methodologies of STEPs 2021 have been published elsewhere¹⁸; therefore, only key components are summarized here. The implementation involved three main STEPs. Step 1 consisted of data collection through a standardized questionnaire¹⁹. Step 2 included physical measurements, such as weight, height, blood pressure, waist circumference, and hip circumference. Step 3 comprised centralized laboratory testing, conducted using the validated Roche-Hitachi Cobas C311 auto-analyzer at the national reference laboratory¹⁸.

Sampling and study population

This study aimed to obtain a nationally representative sample of Iranian adults aged 18 years and older. A systematic cluster random sampling design was employed, proportional to the population distribution across urban and rural areas in all 31 provinces of Iran. A total of 3,176 clusters were defined, each consisting of nine individuals, resulting in an initial sample of 28,821 participants. The sampling framework was stratified by four key demographic variables—age, sex, place of residence (urban/rural), and province—to ensure that the results are representative of the national adult population¹⁸.

Of the initial sample, only participants who had complete data on physical activity and diabetes, based on questionnaire responses, physical measurements, and laboratory tests, and who provided informed consent for blood sampling, were included in the third phase of the survey. This final analytical sample consisted of 18,013 individuals.

The reduction in sample size was primarily due to refusal to participate in blood sampling, incomplete questionnaires, failure to attend follow-up stages, or missing data for key variables. Although such attrition may raise concerns about selection bias, comparative analyses between the initial and final samples revealed no statistically significant differences in the distribution of age, sex, or place of residence. Therefore, the final sample retained the key demographic characteristics of the original population and can reasonably be considered representative of the adult population of Iran for analytical purposes.

Variable definitions

Physical activity assessment

The second version of the Global Physical Activity Questionnaire (GPAQ-2) was used to assess each participant’s weekly physical activity in terms of metabolic equivalent of task (MET) minutes²⁰. Physical activity was assessed in two domains: leisure-time physical activity, defined as moderate- or vigorous-intensity recreational activities, and walking activity, defined as transport-related walking, both excluding occupational activity. The MET unit was assigned as follows: walking time PA at 3.3 MET, moderate PA at 4 MET, and vigorous PA at 8 MET²⁰.

Participants were considered to have sufficient physical activity within a given domain if they achieved ≥ 600 MET-minutes per week in that domain, corresponding to at least 150 min of moderate-intensity, 75 min of vigorous-intensity, or a combination of both²⁰. Those engaging in < 600 MET-minutes per week within a given domain were classified as having IPA in that specific domain.

Previous studies have evaluated the validity and reliability of the GPAQ, demonstrating generally acceptable accuracy and consistency in measurements^21–23.

Diabetes definition

Various variables from the collected survey dataset were utilized to estimate the prevalence of DM, with fasting plasma glucose (FPG) and whole blood hemoglobin A1c (HbA1c) serving as the primary laboratory measures. Whole blood HbA1c was quantified using high-performance liquid chromatography (HPLC), as certified by the reference laboratory. According to the American Diabetes Association’s (ADA) definitions, DM was defined as FPG ≥ 126 mg/dL (7.0 mmol/L) or whole blood Hemoglobin A1c (HbA1c) ≥ 6.5% or the use of oral antihyperglycemic drugs²⁴. This study did not differentiate between type 1 (T1DM) and type 2 diabetes (T2DM).

Glycemic control was defined as HbA1c < 7% among all individuals with diabetes²⁴. Untreated DM was defined as having DM without currently receiving any form of treatment, such as medication or insulin therapy.

Demographic variables

Age was grouped as < 60 or ≥ 60 years, residence as urban or rural, and obesity as BMI ≥ 30 kg/m²¹⁸. Education was classified based on completed years of formal schooling as 0, 1–6, 7–11, and ≥ 12 years.

Calculation of population attributable fraction (PAF)

Conceptual framework

PAF defines the proportion of total cases of a disease in a population that are associated with a particular risk factor and that could theoretically be reduced if that risk factor were completely eliminated, under causal assumptions¹². In this study, we applied Miettinen’s formula to estimate the proportion of DM cases associated with IPA²⁵:

where pc represents the prevalence of insufficient leisure-time or walking activity among individuals with DM in the Iran STEPs 2021 survey, and RR adjusted represents the adjusted relative risk of incident DM associated with insufficient leisure-time or walking activity, derived from external meta-analyses.

The RRs were derived from a comprehensive meta-analysis of prospective cohort studies by Aune et al.²⁶, which examined the relationship between leisure-time and walking activities and the incidence of diabetes in adult populations. The specific RRs applied were:

• Leisure-time inactivity: RR = 1.26 (95% CI 1.20, 1.33) for < 150 min/week versus ≥ 150 min/week of moderate-to-vigorous physical activity²⁶

• Walking inactivity: RR = 1.18 (95% CI 1.10, 1.27) for < 150 min/week versus ≥ 150 min/week of walking²⁶

These RRs were selected because they represent the most comprehensive meta-analysis available, incorporating data from prospective cohort studies with long-term follow-up, thereby minimizing the risk of reverse causality that would affect cross-sectional estimates^1,5,26. This approach is consistent with established methodologies for estimating burden of disease attributable to risk factors, including the Global Burden of Disease studies^1,5,27,28.

PAF measures were calculated separately for sex and age groups (< 60 years and ≥ 60 years) using the same external RRs applied uniformly across subgroups.

The PAFs calculated for leisure-time inactivity and walking inactivity are domain-specific. Because individuals may be inactive in one or both domains (see Supplementary Table S1 for the joint distribution of inactivity across domains), these PAFs are non-additive and cannot be summed to represent a total attributable fraction¹².

To aid interpretation of the PAFs, we estimated internal associations between physical activity domains and diabetes using weighted logistic regression models. Diabetes status was modeled as the dependent variable, and domain-specific physical inactivity (leisure-time inactivity and walking-time inactivity) as independent variables. Three models were fitted: (1) crude, (2) adjusted for age and sex, and (3) multivariable-adjusted models including age, sex, BMI, education level, and place of residence. Adjusted odds ratios (ORs) with 95% CIs are reported in Supplementary Table S2. These internal ORs were not used to calculate PAFs because cross-sectional designs are susceptible to reverse causation and cannot establish temporality; externally derived prospective relative risks are more appropriate for estimating potential causal impact under the assumptions of the PAF framework¹².

Statistical analysis

The demographic parameters of the study population were presented as weighted proportions with corresponding 95% CIs¹⁸. The prevalence of IPA was estimated among diabetic males and females across two age groups (< 60 and ≥ 60 years) using data from the STEPs 2021 survey. The chi-square test was used to assess differences in the prevalence of IPA across demographic variables. All analyses and graphical representations were performed using R for Windows (v4.3.3) and the statistical package “graphPAF”²⁹. Statistical significance was defined as a p-value < 0.05.

Ethics approval and consent to participate

Participants provided written informed consent after being informed of the study’s purpose, procedures, and their right to withdraw. Data were de-identified to ensure confidentiality, with access limited to the principal investigator and database manager. The study was approved by the Research Ethics Committee of the School of Medicine, Tehran University of Medical Sciences (IR.TUMS.MEDICINE.REC. 1403.230) and carried out in accordance with the Declaration of Helsinki.

Results

Participant characteristics and prevalence of insufficient physical activity

We first examined the prevalence of leisure-time and walking-time IPA across key demographic subgroups, stratified by diabetes status (Table 1). Subsequent analyses assessed the association between IPA and DM prevalence, differences in DM treatment and control status by activity level, and the PAF of DM due to IPA, overall and by age and sex.

Table 1.

Prevalence of leisure-time and walking-time inactivity among total, diabetic, and non-diabetic participants in the Iran STEPs survey 2021, stratified by demographic factors.

Variable	Category	Leisure time inactivity		Walking time inactivity
		Percent. (95% CI)	P value*	Percent. (95% CI)	P value*
All participants; n = 18,013
		86.74 (86.29, 87.19(	-	69.95 (69.34, 70.55)	-
Age group	< 60 year	84.7 (84.15, 85.22)	< 0.001	70.1 (69.42, 70.77)	0.343
	≥ 60 year	94.24 (93.54, 94.87)		69.39 (68.06, 70.69)
Sex	Female	92.21 (91.72, 92.68)	< 0.001	74.69 (73.91, 75.45)	< 0.001
	Male	79.97 (79.18, 80.75)		64.08 (63.13, 65.01)
Residency	Urban	86.7 (86.17, 87.21)	0.715	68.04 (66.85, 69.21)	< 0.001
	Rural	86.89 (86, 87.72)		70.58 (69.88, 71.28)
Obesity	Non-obese	85.6 (85.06, 86.12)	< 0.001	68.74 (68.03, 69.44)	< 0.001
	Obese	90.16 (89.34, 90.93)		73.57 (72.39, 74.72)
Education	0	96.47 (95.77, 97.05)	< 0.001	73.97 (72.42, 75.46)	< 0.001
	1–6	92.84 (92.09, 93.51)		69.7 (68.46, 70.91)
	7–11	86.69 (85.63, 87.69)		70.52 (69.12, 71.88)
	12 and over	80.48 (79.68, 81.27)		68.46 (67.51, 69.39)
Participants with diabetes mellitus; n = 3,254, 18.5% (17.9, 19.2)
Age group	< 60 year	91.9 (90.39, 93.3)	0.004	67.9 (65.45, 70.37)	0.001
	≥ 60 year	94.8 (93.41, 95.88)		73.8 (71.29, 76.22)
Sex	Female	95.9 (94.77, 96.74)	< 0.001	74.4 (72.21, 76.56)	< 0.001
	Male	89.5 (87.53, 91.26)		65.3 (62.37, 68.10)
Residency	Urban	92.9 (91.65, 93.95)	0.174	70.5 (68.45, 72.53)	0.744
	Rural	94.4 (92.39, 95.93)		71.2 (67.69, 74.46)
Obesity	Non-obese	92.7 (91.28, 93.89)	0.114	67.3 (64.9, 69.64)	< 0.001
	Obese	94.1 (92.47, 95.36)		75.2 (72.59, 77.71)
Education	0	96.9 (95.44, 97.96)	< 0.001	76.8 (73.63, 79.70)	< 0.001
	1–6	94.8 (93.22, 96.12)		70.9 (67.89, 73.78)
	7–11	92.1 (88.81, 94.44)		69.2 (64.27, 73.73)
	12 and over	87.8 (85, 90.22)		64.4 (60.47, 68.17)
Participants without diabetes mellitus; n = 14,759, 81.5% (80.8, 82.1)
Age group	< 60 year	85.4 (84.69, 86.11)	< 0.001	70.2 (69.29, 71.11)	0.001
	≥ 60 year	94.1 (92.93, 94.96)		65.9 (63.93, 67.94)
Sex	Female	92.2 (91.53, 92.85)	< 0.001	74.3 (73.2, 75.29)	< 0.001
	Male	80.2 (79.12, 81.31)		63.2 (61.89, 64.52)
Residency	Urban	86.4 (85.67, 87.19)	0.010	70.7 (69.69, 71.7)	< 0.001
	Rural	88.1 (87.08, 89.14)		66.7 (65.18, 68.14)
Obesity	Non-obese	86.1 (85.4, 86.84)	0.001	68.3 (67.37, 69.29)	< 0.001
	Obese	89.6 (88.44, 90.73)		72.8 (71.15, 74.43)
Education	0	96.2 (95.15, 96.96)	< 0.001	70.7 (68.52, 72.73)	0.060
	1–6	91.7 (90.61, 92.6)		67.5 (65.82, 69.11)
	7–11	86.9 (85.48, 88.2)		70.1 (68.25, 71.93)
	12 and over	80.6 (85, 90.2)		69.7 (68.33, 71.01)

*P-values represent the comparison of inactivity prevalence across different levels of demographic variables, as determined by the chi-square test. CI Confidence Interval.

Among participants with DM, leisure-time inactivity was significantly higher among older adults aged ≥ 60 years (94.8%, 95% CI 93.41, 95.88) compared to those aged < 60 years (91.9%, 95% CI 90.39, 93.3; P = 0.004). Similarly, walking-time inactivity was significantly higher among older diabetic adults (73.8%, 95% CI 71.29, 76.22) compared to younger diabetic adults (67.9%, 95% CI 65.45, 70.37; P = 0.001). Among non-diabetic participants, leisure-time inactivity was also higher in older adults (94.1%, 95% CI 92.93, 94.96) compared to younger adults (85.4%, 95% CI 84.69, 86.11; P < 0.001). However, walking-time inactivity among non-diabetic participants was lower in older adults (65.9%, 95% CI 63.93, 67.94) compared to younger adults (70.2%, 95% CI 69.29, 71.11; P = 0.001).

Female participants showed significantly higher inactivity prevalence compared to males across both domains and diabetes status. Among individuals with DM, leisure-time inactivity was 95.9% (95% CI 94.77, 96.74) in females versus 89.5% (95% CI 87.53, 91.26) in males (P < 0.001), and walking-time inactivity was 74.4% (95% CI 72.21, 76.56) in females versus 65.3% (95% CI 62.37, 68.10) in males (P < 0.001). Similar patterns were observed among non-diabetic participants for both leisure-time (92.2% vs. 80.2%, P < 0.001) and walking-time inactivity (74.3% vs. 63.2%, P < 0.001).

Obesity was significantly associated with higher walking-time inactivity prevalence among both diabetic (75.2%, 95% CI 72.59, 77.71 vs. 67.3%, 95% CI 64.9, 69.64; P < 0.001) and non-diabetic participants (72.8%, 95% CI 71.15, 74.43 vs. 68.3%, 95% CI 67.37, 69.29; P < 0.001). Leisure-time inactivity was slightly higher in obese individuals with DM (94.1%, 95% CI 92.47, 95.36) compared to non-obese individuals with DM (92.7%, 95% CI 91.28, 93.89), but this difference was not statistically significant (P = 0.114).

Education level showed a consistent inverse relationship with leisure-time inactivity among non-diabetic participants. The prevalence of leisure-time inactivity decreased progressively with increasing education: 96.2% (95% CI 95.15, 96.96) among those with no formal education, 93.4% (95% CI 92.59, 94.20) among those with 1–6 years, 88.9% (95% CI 87.89, 89.80) among those with 7–11 years, and 80.6% (95% CI 79.68, 81.27) among those with ≥ 12 years of education (P for trend < 0.001). Among non-diabetic participants, leisure-time inactivity was slightly higher among rural residents (88.1%, 95% CI: 87.08, 89.14) compared to urban residents (86.4%, 95% CI 85.67, 87.19; P = 0.010).

Association between physical activity and diabetes prevalence

Table 2 presents the comparison of DM prevalence between active and inactive individuals in the leisure-time and walking domains. Among individuals who were inactive during leisure time, 19.61% (95% CI 18.92, 20.32) had DM, compared to 10.5% (95% CI 9.12, 12.07) among those who were active (P < 0.001).

Table 2.

Comparison of diabetes status across active and inactive individuals in leisure and walking activities in the Iran STEPs survey 2021.

	Leisure time activity		Walking time activity
Total population, n (%)	Inactive	Active	Inactive	Active
	Percent. (95% CI)		Percent. (95% CI)
No diabetes mellitus: 14,759 (81.49)	80.39 (79.68, 81.08)	89.5 (87.93, 90.88)	81.19 (80.41, 81.96)	82.1 (80.93, 83.22)
Diabetes: 3,254 (18.51)	19.61 (18.92, 20.32)	10.5 (9.117, 12.07)	18.81 (18.04, 19.59)	17.9 (16.78, 19.07)
P value *	< 0.001		0.017
Diabetes mellitus subgroup, n (%)
Untreated diabetes mellitus: 1,639 (50.2)	49.94 (47.95, 51.93)	53.54 (46.1, 60.83)	49.41 (47.12, 51.7)	52.04 (48.49, 55.57)
Uncontrolled diabetes mellitus: 961 (30.4)	30.42 (28.6, 32.29)	30.37 (23.96, 37.65)	30.86 (28.77, 33.03)	29.34 (26.21, 32.68)
Controlled diabetes mellitus: 634 (19.4)	19.64 (18.11, 21.27)	16.09 (11.37, 22.29)	19.73 (17.97, 21.61)	18.62 (16.01, 21.54)
P value **	0.479		0.475

*P values refer to the comparison between diabetes status (diabetes vs. no diabetes) across activity levels (inactive vs. active) in the total study population, using the chi-square test.

**P values refer to the comparison of diabetes treatment and glycemic control status (untreated, uncontrolled, controlled) across activity levels (inactive vs. active) within the subgroup of participants with diabetes mellitus, using the chi-square test.

Figure 1 illustrates the prevalence of DM across increasing levels of physical activity, stratified by sex. An inverse association between DM prevalence and activity level was clearly observed for leisure-time activity, while patterns for walking activity were comparatively flat.

Fig. 1.

Prevalence of diabetes by sex and physical activity (PA) levels in the Iran STEPs Survey 2021. (A) Prevalence of diabetes by sex and leisure-time PA level, categorized by METs (Metabolic Equivalent of Tasks) minutes per week: Less than 600, 600–3000, and more than 3000. (B) Prevalence of diabetes by sex and walking-time PA level, categorized by METs minutes per week: Less than 600, 600–3000, and more than 3000. Lines represent prevalence trends across different PA levels for females (blue), males (orange), and the overall population (green).

In contrast, DM prevalence differed only modestly by walking activity status, with 18.81% (95% CI 18.04, 19.59) among inactive individuals versus 17.9% (95% CI 16.78, 19.07) among active individuals (P = 0.017).

To formally evaluate the internal association between walking inactivity and diabetes in this cross-sectional sample, we conducted weighted logistic regression analyses (Supplementary Table S2). For walking inactivity, the internally estimated association was modest in magnitude, with an adjusted odds ratio of 1.11 (95% CI 1.02, 1.21) in age- and sex-adjusted models and 1.09 (95% CI 1.01, 1.19) in multivariable-adjusted models (Supplementary Table S2). In contrast, leisure-time inactivity showed a stronger internal association (multivariable-adjusted OR = 1.55; 95% CI 1.32, 1.82). As detailed in the Methods, these internal ORs were not used for PAF calculation due to the cross-sectional design’s susceptibility to reverse causation.

Diabetes treatment and glycemic control by activity status

Among participants with DM, the distribution of treatment and glycemic control status did not differ significantly by activity level. For leisure-time activity, inactive individuals had 49.94% (95% CI 47.95, 51.93) untreated, 30.42% (95% CI 28.60, 32.29) uncontrolled (treated but HbA1c ≥ 7%), and 19.64% (95% CI 18.11, 21.27) controlled (treated with HbA1c < 7%). Active individuals showed similar proportions: 53.54% (95% CI 46.10, 60.83) untreated, 30.37% (95% CI 23.96, 37.65) uncontrolled, and 16.09% (95% CI 11.37, 22.29) controlled (P = 0.479).

Notably, approximately half of all individuals with DM remained untreated regardless of their activity level (49.9% among inactive and 53.5% among active participants).

Population attributable fraction estimates

Table 3 presents the PAF of DM associated with domain-specific IPA. In the total population, 14.6% (95% CI 11.5, 17.6) of DM cases were associated with leisure-time inactivity, while 15.7% (95% CI 12.4, 18.9) were associated with walking-time inactivity. These estimates are domain-specific and non-additive.

Table 3.

Population attributable fraction of leisure time and walking inactivity for diabetes in the Iran STEPs survey 2021, stratified by age and sex.

	PAF
	Total population		Males		Females
	Leisure time	Walking time	Leisure time	Walking time	Leisure time	Walking time
Age group	Percent. (95% CI)		Percent. (95% CI)		Percent. (95% CI)
< 60 years	14.5 (11.4, 17.4)	15.2 (11.9, 18.3)	13.8 (10.8, 16.7)	14.8 (11.6, 18.0)	14.8 (11.7, 17.9)	15.4 (12.1, 18.6)
≥ 60 years	14.8 (11.7, 17.9)	16.3 (12.8, 19.6)	14.4 (11.3, 17.4)	14.5 (11.3, 17.6)	15.1 (11.9, 18.2)	17.5 (13.9, 21.1)
Total	14.6 (11.5, 17.6)	15.7 (12.4, 18.9)	14.1 (11.1 ,17.1)	14.6 (11.5, 17.8)	15 (11.8, 18.0)	16.4 (12.9, 19.7)

PAF: Population attributable fraction, CI: confidence interval.

Despite similar PAF magnitudes, the underlying internal associations differ between domains. The leisure-time PAF was supported by a multivariable-adjusted internal odds ratio of 1.55 (95% CI 1.32, 1.82) and an inactivity prevalence of 86.7% in the total population. The walking PAF was supported by a multivariable-adjusted internal odds ratio of 1.09 (95% CI 1.01, 1.19; Supplementary Table S2), an inactivity prevalence of 70.0%, and an external relative risk of 1.18 (95% CI 1.10, 1.27)²⁶.

When stratified by age, PAFs were slightly higher among individuals aged ≥ 60 years compared to those < 60 years. For leisure-time inactivity, PAF was 14.8% (95% CI 11.7, 17.9) in older adults versus 14.5% (95% CI 11.4, 17.4) in younger adults. For walking inactivity, PAF was 16.3% (95% CI 12.8, 19.6) versus 15.2% (95% CI 11.9, 18.3), respectively.

Sex-specific analysis showed higher estimated PAFs among females compared to males. Among females, PAFs were 15.0% (95% CI 11.8, 18.0) for leisure-time inactivity and 16.4% (95% CI 12.9, 19.7) for walking inactivity. Among males, corresponding estimates were 14.1% (95% CI 11.1, 17.1) and 14.6% (95% CI 11.5, 17.8).

As shown in Supplementary Table S1, 63.55% of the total population and 67.05% of individuals with diabetes were inactive in both domains.

Discussion

This nationwide cross-sectional study analyzed Iran STEPs-2021 data to estimate the population attributable fraction (PAF) of DM associated with IPA in both leisure-time and walking domains. We found that approximately 15% of DM cases could be associated with IPA under causal assumptions, with higher estimated burdens observed among women, older adults, and individuals with lower educational attainment. We also observed an alarmingly high overall prevalence of leisure-time and walking-time inactivity (86.7% and 70.0%, respectively). IPA was more common among older adults, females, individuals with obesity, and those with lower education levels. Among participants with DM, IPA remained markedly high, with older age and female sex again associated with greater inactivity. Notably, DM prevalence showed a strong association with insufficient leisure-time activity in the internal cross-sectional analyses, whereas the association with insufficient walking-time activity was modest. Moreover, we observed no significant differences in DM treatment or glycemic control status between physically active and inactive individuals. Importantly, the domain-specific PAFs reported are non-additive and should not be summed, as individuals may be inactive in one or both domains¹².

Acknowledging IPA as the fourth leading risk factor for various NCDs and premature mortality, the WHO has set a target of a 15% relative reduction in global IPA prevalence by 2030 as part of broader initiatives to improve NCD prevention and control³⁰. Our finding that approximately 15% of DM cases could be hypothetically associated with leisure-time inactivity aligns closely with this target, suggesting that achieving the WHO global goals could yield meaningful population-level diabetes prevention benefits in Iran. However, as emphasized throughout this manuscript, these estimates represent hypothetical potential impact fractions under strong causal assumptions rather than precise predictions of preventable cases¹².

A study based on the GBD 2019 data found that the age-standardized rates of DALYs and deaths due to DM attributable to IPA showed significant upward trends from 1990 to 2019 ²⁷. The PAF estimates for DM attributable to IPA have varied widely across studies. A prior systematic review reported that the PAF of DM due to IPA among individuals who never engaged in strenuous sports was 13% (95% CI 3, 22) in males and 29% (95% CI 17, 41) in females¹⁴. Our estimates for Iran approximately 17% for leisure-time inactivity in the total population, with higher estimates among women are broadly consistent with these international comparisons, though direct comparability is limited by differences in exposure definition, outcome ascertainment, and population characteristics.

The relatively weak internal association between walking-time inactivity and diabetes prevalence in the cross-sectional data, despite a non-trivial calculated population attributable fraction (PAF) of 15.7%, requires careful interpretation. This apparent discrepancy reflects the mathematical structure of Miettinen’s formula, in which a high prevalence of exposure combined with an externally derived relative risk can yield a non-zero PAF even when internal associations are weak^12,14. This does not invalidate the PAF approach but rather highlights the importance of interpreting PAFs within their methodological context. Because cross-sectional designs are susceptible to reverse causation and cannot establish temporality, externally derived prospective relative risks are more appropriate for estimating potential causal impact¹².

Several explanations may account for the modest internal association observed for walking inactivity. First, most walking reported in the Iran STEPs 2021 survey was likely light and incidental, falling short of the ≥ 3 METs intensity generally required to improve insulin sensitivity. Evidence indicates that only brisk or moderate-to-vigorous walking meaningfully reduces T2DM risk, whereas slow-paced walking offers little protective benefit^31–34. A recent meta-analysis demonstrated that walking speed is a critical determinant of diabetes risk reduction, with brisk walkers showing substantially lower risk compared to slow walkers³⁴. Second, the GPAQ shows only poor to fair agreement with accelerometer data²¹, with the greatest misclassification occurring for moderate-intensity activities³⁵. Together, insufficient activity intensity and imprecise measurement may have obscured any potential protective effect in the cross-sectional analysis. This highlights the importance of not interpreting cross-sectional associations as direct evidence of causal effects and underscores the need for cautious interpretation of domain-specific PAFs when internal associations are null^12,14.

Several complementary mechanisms may explain our unexpected finding that glycemic treatment and control did not differ between “active” and “inactive” respondents. This descriptive finding should be interpreted cautiously and not as evidence that physical activity has limited relevance for diabetes management.

First, the GPAQ captures only the past seven days of activity and does not differentiate structured exercise from incidental movement, likely leading to misclassification of habitual activity levels. This is particularly relevant because randomized trials demonstrate that sustained moderate-to-vigorous exercise of ≥ 150 min per week reduces HbA1c by 0.4–0.6 percentage points^36,37, but such benefits require consistent engagement over time. Second, glycemic indices are strongly influenced by factors not measured in the STEPs dataset, including diabetes duration, medication regimen and treatment intensification, and medication adherence^38,39. Cohort data indicate that while over 90% of patients adhere to their medications, fewer than 25% meet recommended exercise guidelines³⁸. These unmeasured factors could have masked or confounded any relationship between activity and glycemic control.

Third, reverse causation is plausible: individuals with diabetes and poor glycemic control may have diabetes-related complications that limit their ability to engage in physical activity^40,41. This aligns with the “physical activity paradox” literature, which suggests that the relationship between activity and health outcomes may be bidirectional, particularly in populations with existing disease^42,43. Peripheral neuropathy, cardiovascular disease, and obesity are all more common in poorly controlled diabetes and independently predict lower physical activity levels^40,44–46. A cross-sectional analysis of 6,856 adults with DM revealed that comorbidities, higher BMI, and poorer quality of life collectively reduced the likelihood of meeting guideline-recommended levels of moderate-to-vigorous physical activity by approximately 40%⁴⁶. Thus, reverse causation, measurement limitations, and unmeasured confounding together explain the null findings, which should not be interpreted as evidence that physical activity is irrelevant for diabetes management.

In the present study, we observed more pronounced PAFs for IPA among females and older adults. Previous studies have also indicated significant sex- and age-based disparities in the burden of diseases attributable to IPA in the MENA region^13,28. A recent analysis of the GBD 2021 data for the MENA region found that females had higher age-standardized DALY rates attributable to low physical activity compared to males¹³.

In the MENA region, females’ participation in physical activity is often hindered by sociocultural norms, traditional sex roles, and conflicting responsibilities⁴⁷. To reduce these disparities and support equitable health outcomes, tailored interventions such as females-only fitness spaces, community-based programs delivered in trusted settings, and programs designed for specific life stages may be helpful^3,47. The elevated PAFs for IPA among older adults may reflect age-associated declines in mobility and physical function, which limit their ability to maintain regular activity, as well as the cumulative effects of reverse causation whereby diabetes complications further reduce activity capacity^44–46.

Consistent with our findings, previous research has consistently shown that IPA is more prevalent among individuals with lower income and educational attainment^48,49. This pattern may be explained by the tendency of individuals with higher socioeconomic status (SES) to engage more in leisure-time physical activity, while those with lower SES often face constraints in resources, time, or access that limit participation in leisure-time or other forms of PA^50,51. Additionally, occupational physical activity which is more common among lower SES groups was excluded from our definition, potentially underestimating total activity in these populations^2,42.

Strengths and limitations

This study leverages the strengths of the Iran STEPS 2021 survey, including its large nationally representative sample, standardized WHO protocols, and objective biomarkers (fasting plasma glucose and HbA1c) for diabetes diagnosis rather than relying solely on self-reported status, which substantially enhances the validity of prevalence estimates^17,18. The large sample size enables stratified analyses by key demographic subgroups to identify populations with differential attributable burdens.

The PAF estimates presented in this study should be interpreted as domain-specific, non-additive, approximate hypothetical impact fractions under strong causal assumptions, most appropriately used for comparative purposes rather than as precise predictions of preventable cases¹².

Several important limitations warrant consideration. First, we defined exposure using leisure-time physical activity and transport-related walking, excluding occupational activity entirely. In low- and middle-income countries such as Iran, where manual labor constitutes a substantial proportion of total energy expenditure, this exclusion may misclassify individuals engaged in physically demanding occupations as “insufficiently active,” potentially inflating the estimated prevalence of IPA and affecting PAF calculations^2,42. However, given the “physical activity paradox” whereby occupational activity may not confer the same cardiometabolic benefits as leisure-time activity, our estimates should be interpreted as attributable to low leisure-time activity and low transport walking specifically^42,43. Furthermore, the domain-specific threshold approach differs from standard WHO methodology, which typically considers the sum of activity across all domains²⁰. The observed overlap between leisure-time and walking inactivity (Supplementary Table S1) demonstrates why domain-specific PAFs are non-additive and should not be interpreted as independent or summable contributions¹².

Second, the use of self-reported GPAQ data introduces potential recall and social desirability bias, which may lead to overestimation of physical activity levels ^21,22. Studies comparing GPAQ with accelerometer-measured activity have demonstrated limited concurrent validity for moderate-intensity activities, with a tendency for over-reporting ^23,35. If the walking domain captured in Iran STEPs predominantly reflects low-intensity or incidental walking, this concern is particularly relevant for the walking-related PAF.

Third, our PAF estimates depend on relative risks derived primarily from non-Iranian prospective cohorts, as Iran-specific longitudinal RRs for leisure-time activity and transport-related walking are scarce ²⁶. The meta-analysis by Aune et al. predominantly included studies from high-income countries, with no studies from Iran or the broader MENA region²⁶. The applicability of these RRs to the Iranian population rests on the assumption that the biological effect of physical activity on diabetes risk is transportable across populations—an assumption that may not hold given differences in activity patterns, walking intensity, baseline metabolic risk, and healthcare access^13,34,52,53. If the walking domain captured in Iran STEPs reflects lower-intensity activity compared with the cohorts underpinning the meta-analysis—a concern supported by prospective data indicating substantially lower diabetes risk among brisk versus slow walkers^31,33,34 the walking-related PAF may be overestimated. In the presence of measurement error or the inclusion of low-intensity walking, PAFs calculated using cohort-derived RRs likely represent upper bound estimates rather than internally validated attributable fractions.

Fourth, the use of uniform RRs across age and sex subgroups, due to the absence of subgroup-specific estimates, may have led to over- or underestimation. Prospective studies indicate that the magnitude of the physical activity–type 2 diabetes association varies by sex, age, and adiposity^52–54. Because PAF is a monotonic function of the RR, even modest departures of the true subgroup-specific RR from the pooled estimate can translate into meaningful differences in subgroup PAFs when the prevalence of inactivity among cases is high¹². This mathematical constraint applies to PAF interpretation generally, not only to these subgroups.

Fifth, our reported relative risks were not adjusted for all potential confounders, notably obesity, which can act as both a mediator and a confounder in the physical activity–diabetes relationship⁵⁴. Consequently, our estimates represent the combined direct and indirect effects of physical inactivity, which from a policy perspective captures the total potential benefit of reducing inactivity but precludes disentangling the independent effect of physical activity separate from its impact on adiposity^1,5.

Sixth, the cross-sectional design fundamentally limits causal inference and precludes firm conclusions about directionality¹². Reverse causation is a particular concern: individuals with diabetes may reduce their physical activity due to disease complications or following diagnosis, which would increase the observed prevalence of IPA among diabetics and potentially inflate PAF estimates^{40,41,44–46}. The finding that walking inactivity was associated with a modest increase in diabetes prevalence in the internal cross-sectional data, yet yields a substantial PAF of 15.7%, illustrates this paradox and underscores the importance of not interpreting cross-sectional associations as direct evidence of causal effects^12,14.

Seventh, the study assesses physical activity based solely on volume without independently examining intensity. There is strong evidence that moderate-to-vigorous physical activity provides a stronger protective effect against DM than light-intensity activity^32–34. By using a fixed threshold that combines intensity and duration, we may have missed important variation in the quality of activity, particularly among older adults and women who may accumulate activity at lower intensities⁵³.

Eighth, the finding that physical activity level was not associated with diabetes treatment or glycemic control should be interpreted cautiously due to measurement limitations and unmeasured confounding^36,38,39. The GPAQ captures short-term activity and does not assess structured exercise dose or long-term patterns²¹. Moreover, the survey does not capture diabetes duration, medication type, treatment intensification, or medication adherence all strong determinants of glycemic control^36,38,39. These results should be viewed as descriptive rather than inferential.

Ninth, the study reports IPA prevalence approaching or exceeding 95% in high-risk subgroups such as women and those with no formal education. When the prevalence of the risk factor among cases approaches 100%, the PAF asymptotically approaches a maximum value determined solely by the RR¹². Consequently, the preventive potential through increasing physical activity alone in these groups is mathematically constrained by risk factor saturation, highlighting the need to complement physical activity promotion with secondary prevention and addressing other modifiable factors such as obesity and diet⁵⁵.

These limitations underscore the need for future longitudinal studies in the MENA region, device-based physical activity measurement, and more granular risk modeling to refine estimates and better inform policy^30,56. Despite these caveats, the consistency of our findings with global burden of disease estimates^1,5,13,28 and the large representative sample strengthen confidence in the overall conclusion that IPA particularly in leisure time contributes substantially to the diabetes burden in Iran..

Policy implications and future directions

Despite methodological limitations, our findings reinforce insufficient physical activity (IPA) as an important modifiable contributor to the diabetes burden in Iran. The high prevalence of inactivity among women, older adults, and disadvantaged groups highlights persistent structural barriers requiring targeted policy responses. The estimated 15% population attributable fraction for leisure-time inactivity aligns with the WHO Global Action Plan target of a 15% relative reduction in global IPA prevalence by 2030³⁰, suggesting that achieving these global targets could yield meaningful diabetes prevention benefits in Iran. However, because inactivity prevalence approaches universality in several subgroups for example, 96% among adults with no formal education and 96% among women with diabetes these PAFs represent upper-bound counterfactual estimates rather than realistic short-term gains. When risk factor prevalence approaches saturation, even large increases in physical activity yield modest relative reductions in disease burden¹². In such contexts, physical activity promotion must be embedded within comprehensive strategies addressing obesity, unhealthy diets, and equitable healthcare access^5,55.

Achieving sustainable increases in physical activity requires multisectoral collaboration across health, transport, urban planning, education, and municipal sectors³⁰. Built environment interventions—including pedestrian infrastructure, green spaces, and safe transport systems—can facilitate routine walking, particularly in underserved areas^56,57. However, evidence from low- and middle-income countries remains limited, underscoring the need for context-specific evaluation^56,58.

Equity-oriented interventions should prioritize gender-responsive approaches such as women-only fitness facilities and community-based programs delivered in trusted settings^3,47, age-adapted services integrated into primary care⁵⁷, and programs addressing cost and access barriers for disadvantaged groups^50,51. Within healthcare, routine physical activity assessment and brief counseling should be strengthened in diabetes prevention pathways, with structured lifestyle interventions modeled on successful trials^6–8. However, health system constraints in Iran including limited primary care infrastructure for preventive counseling require task-sharing, training, and integration with existing non-communicable disease screening programs⁵⁵.

Future research should prioritize longitudinal cohort studies in Iran and the MENA region incorporating device-based physical activity measurement to strengthen causal inference and reduce exposure misclassification^23,35. Context-specific risk estimates stratified by age, sex, and adiposity are urgently needed to refine PAF calculations^52–54. Implementation research evaluating the effectiveness and cost-effectiveness of context-appropriate interventions will be critical for evidence-based policymaking^30,58. Health economic evaluations are also needed to build investment cases for physical activity interventions in Iran. Finally, strengthening national surveillance systems and investing in prospective research infrastructure will be essential for monitoring progress and supporting evidence-based diabetes prevention policies.

Conclusion

This nationally representative analysis estimated that approximately 15% of prevalent diabetes cases in Iran could be associated with leisure-time physical inactivity under causal assumptions and assuming transportability of external relative risks to the Iranian population, with higher estimated burdens among women, older adults, and less-educated groups.

The modest internal association observed between walking inactivity and diabetes prevalence—in contrast to the stronger association for leisure-time inactivity likely reflects the limitations of cross-sectional data, self-reported measurement, and the critical importance of activity intensity, specifically that moderate-to-vigorous rather than light-intensity walking is required to confer metabolic benefit. Similarly, the null findings for glycemic control should not be interpreted as evidence that physical activity is irrelevant for diabetes management, but rather reflect the limitations of short-term self-reported activity and unmeasured confounding by diabetes duration, medication adherence, and complications.

If these associations reflect modifiable causal relationships, and if the external relative risks are transportable to the Iranian population, increasing physical activity—particularly leisure-time and moderate-to-vigorous activity—could contribute to lower diabetes prevalence. However, the near-universal inactivity prevalence in several subgroups, such as 96% among adults with no formal education and 96% among women with diabetes, suggests potential risk factor saturation. This indicates that population attributable fractions represent upper-bound counterfactual estimates rather than achievable short-term gains from activity promotion alone. In these contexts, physical activity promotion must be embedded within comprehensive strategies addressing obesity, diet, and equitable healthcare access.

Achieving meaningful reductions in diabetes burden will require multisectoral collaboration to create supportive environments for active lifestyles, alongside targeted interventions addressing barriers faced by women, older adults, and disadvantaged groups. Future longitudinal studies with device-based physical activity measurement are needed to strengthen causal inference and guide effective public health policies in Iran.

Supplementary Information

Supplementary Information. ^{(20.3KB, docx)}

Abbreviations

IPA

Insufficient physical activity

DM

Diabetes melitus

DALY

Disability-adjusted life year

MENA

Middle East and North Africa

MET

Metabolic equivalent of task

PAF

Population-attributable fraction

RR

Relative risk

STEPS

STEPwise approach to surveillance of non-communicable disease risk factors

GPAQ

Global physical activity questionnaire

WHO

World health organization

NCD

Non-communicable disease

LMICs

Low- and middle-income countries

SE

Standard error

SES

Socioeconomic status

CI

Confidence interval

OR

Odds ratio

PA

Physical activity

UI

Uncertainty interval

Author contributions

Navid Ebrahimi: Writing—Original Draft, Writing—Review & Editing. Keyvan Karimi: Methodology, Formal Analysis, Writing—Original Draft. Samaneh Asgari: Methodology, Validation,Writing—Review & Editing. Sepehr Khosravi: Conceptualization, Writing—Review & Editing. Sina Azadnajafabad: Writing—Review & Editing. Yosra Azizpour: Writing—Review & Editing. Ali Golestani: Conceptualization, Writing—Review & Editing. Elham Ahmadnezhad: Writing—Review & Editing. Afshin Ostovar: Writing—Review & Editing. Hanieh-Sadat Ejtahed: Writing—Review & Editing. Maryam Selk-Ghaffari: Writing—Review & Editing. Moloud Payab, Funding Acquisition: Writing—Review & Editing. Amirhossein Takian: Writing—Review & Editing. Rajabali Daroudi: Writing—Review & Editing. Samaneh Akbarpour: Conceptualization, Methodology, Validation, Writing—Review & Editing, Supervision. Nazila Rezaei: Conceptualization, Methodology, writing—Review & Editing, Supervision, All authors read and approved the final version of the manuscript and agree to be accountable for all aspects of the work.

Funding

This work was financially supported by Tehran University of Medical Sciences, Tehran, Iran (Grant number: 1402–4-221–69745). The funding organization had no involvement in the design of the study, data collection, data analysis, interpretation of findings, decision to publish, or preparation of the manuscript.

Data availability

The datasets generated or analyzed in this study are not publicly accessible due to contractual obligations with the funding agency, the National Institute of Health Research of the Islamic Republic of Iran. Nevertheless, data may be made available upon reasonable request to the corresponding author, subject to the funding agency’s terms and conditions.

Declarations

Competing interests

The authors declare no competing interests.