Sarcopenia in Indian adults with type 2 diabetes: a systematic review and meta-analysis of prevalence and associated factors

Abstract

Objectives

To determine the pooled prevalence of sarcopenia phenotypes and synthesise the evidence on factors associated with sarcopenia among adults with type 2 diabetes mellitus (T2DM) in India.

Design

Systematic review and meta-analysis.

Data sources

PubMed, Embase, Scopus and Web of Science were searched from inception to 14 January 2026, supplemented by reference list and citation searching.

Eligibility criteria for selecting studies

Cross-sectional studies of Indian adults (≥18 years) with T2DM reporting sarcopenia prevalence using Asian Working Group for Sarcopenia (AWGS) diagnostic criteria.

Data extraction and synthesis

Two reviewers independently screened and extracted data. Methodological quality was assessed using the Joanna Briggs Institute prevalence checklist. Proportions were logit-transformed and pooled using random-effects models with Hartung-Knapp CIs and prediction intervals. Associated factors were synthesised narratively within a tiered framework separating adjusted from unadjusted ORs.

Results

Seven studies (n=1359) were included. Pooled confirmed sarcopenia prevalence (AWGS 2019, k=6) was 27.4% (95% Hartung-Knapp CI 18.9% to 37.9%) but highly imprecise: I²=91.7% and the 95% prediction interval ran from 6.9% to 65.6%. Subgroup analysis showed 20.2% for dual-energy X-ray absorptiometry versus 31.7% for bioelectrical impedance analysis (Qbetween p=0.011). Severe sarcopenia was 11.0% (k=5) and sarcopenic obesity 25.2% (k=4); possible sarcopenia (65.6%, k=3; I²=98.6%) was too heterogeneous to interpret. In the two studies reporting adjusted estimates, hypertension (both studies) and diabetes duration ≥10 years (one study) were positively but imprecisely associated with sarcopenia. Certainty was very low for all prevalence outcomes (Grading of Recommendations Assessment, Development and Evaluation (GRADE)).

Conclusions

Sarcopenia appears common among Indian adults with T2DM, at approximately one in four, but certainty is very low and the plausible prevalence in a new cohort spans roughly 7% to 66%. These estimates are not a sufficient basis for a screening recommendation, but justify greater clinical awareness and, above all, community-based, dual-energy X-ray absorptiometry-standardised prospective studies, including from South India.

PROSPERO registration number

CRD420251251024.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• This systematic review and meta-analysis was prospectively registered in the International Prospective Register of Systematic Reviews (PROSPERO) and conducted with dual independent screening and data extraction verified against source publications, reported per Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020, PRISMA-S and Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines.

• A tiered synthesis framework explicitly separated adjusted from unadjusted ORs, with factors associated with sarcopenia synthesised narratively because too few studies (two per factor) were available for reliable pooling.

• Methodological quality was rigorously assessed using the Joanna Briggs Institute prevalence checklist with ratings verified against source publications.

• The small number of eligible studies limits the power of subgroup analyses, and substantial between-study heterogeneity means the pooled estimates are imprecise and must be interpreted alongside their prediction intervals and Grading of Recommendations Assessment, Development and Evaluation (GRADE) certainty rating.

• All included studies recruited from tertiary care or camp settings, and no studies were available from South India, limiting generalisability to community-dwelling and geographically diverse populations.

Introduction

India harbours approximately 101 million adults with type 2 diabetes mellitus (T2DM), the largest absolute burden of any nation.^{1 2} Chronic hyperglycaemia drives skeletal muscle catabolism through converging mechanisms including insulin resistance-mediated impairment of protein synthesis, ubiquitin-proteasome pathway activation, mitochondrial dysfunction and low-grade systemic inflammation, collectively accelerating the age-related loss of muscle mass, strength and function.^3–5 The resulting deterioration compounds the metabolic burden of T2DM and is increasingly recognised as a clinically actionable target.

Sarcopenia is defined by the Asian Working Group for Sarcopenia (AWGS 2019) consensus as low appendicular skeletal muscle mass index, measured by dual-energy X-ray absorptiometry (DXA) or bioelectrical impedance analysis (BIA), together with reduced handgrip strength or diminished physical performance.⁶ AWGS 2019 delineates three phenotypes—possible sarcopenia (low strength or poor performance alone), confirmed sarcopenia (low muscle mass plus low strength or poor performance) and severe sarcopenia (all three criteria met)—and additionally recognises sarcopenic obesity as a particularly adverse combination.⁶

India presents a distinctive epidemiological context because of the widely documented ‘thin-fat’ phenotype—disproportionately low lean muscle mass coexisting with high relative adiposity at body mass indices considered normal by conventional Western thresholds.^{7 8} Compounded by lifelong dietary protein insufficiency, widespread vitamin D deficiency and sedentary occupational patterns, this phenotype plausibly places Indian adults with T2DM at higher sarcopenia risk than their counterparts in East Asia or Europe.^{8 9} Indian-specific data are therefore essential to guide clinically appropriate screening thresholds.

Globally, sarcopenia prevalence in adults with T2DM ranges from approximately 15% to 44%, with a meta-analytic estimate of roughly 18% across predominantly Asian and European cohorts,¹⁰ while a meta-analysis of community-dwelling Asian adults aged ≥60 years reported 15.9% among those with diabetes versus 10.8% among those without.¹¹ Two recent Asia-wide syntheses have since enlarged this evidence base. Wu et al pooled 43 studies (19 055 participants) of adults aged ≥45 years across Asia and reported an overall prevalence of 17% (95% CI 14% to 21%), identifying older age, higher HbA1c, diabetic nephropathy and neuropathy and higher body fat percentage as associated factors and higher body mass index, better nutritional status and metformin use as protective.¹² Yogesh et al restricted inclusion to AWGS 2019 studies and pooled 39 studies (approximately 19 902 participants), reporting 23% (95% CI 18% to 27%) for confirmed sarcopenia, 61% for possible sarcopenia and 12.1% for severe sarcopenia, with older age, male sex, hypertension, longer diabetes duration and physical inactivity among the associated factors.¹³

Neither review, however, was designed to yield an estimate Indian clinicians and policymakers can apply. Indian studies contributed 2 of 43 studies (388 participants) to Wu et al¹² and 5 of 39 to Yogesh et al, and in both were analysed as one stratum of a wider regional subgroup. Both nevertheless signalled that the Indian estimate may depart from the pan-Asian average: Wu et al reported an India subgroup prevalence of 20% (95% CI 16% to 24%), and the five studies constituting the higher South-East Asia subgroup of Yogesh et al (37.5%) were all conducted in India. Neither characterised the full AWGS phenotypic spectrum in an Indian population, applied the prediction intervals (PIs) and small-sample variance corrections appropriate to a small evidence base or examined the India-specific determinants—the ‘thin-fat’ phenotype, habitually low-quality protein intake and tertiary-care recruitment—that plausibly shape the Indian estimate and its heterogeneity. Indian studies applying AWGS criteria report prevalences spanning approximately 19% to over 60%,^14–20 and this fragmented evidence has not previously been synthesised in its own right. The present review addresses that gap rather than duplicating the existing pan-Asian syntheses.

This review aimed to (i) determine the pooled prevalence of confirmed sarcopenia and characterise the phenotypic spectrum—including possible sarcopenia, severe sarcopenia and sarcopenic obesity—among adults with T2DM in India; and (ii) synthesise the evidence on factors associated with sarcopenia, including age, diabetes duration, glycaemic control, hypertension, neuropathy and physical inactivity.

Methods

Protocol and reporting standards

This systematic review and meta-analysis was prospectively registered in PROSPERO (CRD420251251024; registered 24 December 2025) and is reported per the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 statement²¹ and the Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines.²² The registration was subsequently updated (V.1.1, 25 February 2026); the full version history is available in the PROSPERO record. The registered title differs from the title of this report (online supplemental note S1). The literature search is additionally reported in accordance with the PRISMA-S extension for reporting literature searches in systematic reviews.²³ The completed PRISMA 2020, MOOSE and PRISMA-S checklists are provided in online supplemental appendices S1,S2 and S3.

Review question (PICOT framework)

The review question and selection criteria were structured using the PICOT (Population, Intervention, Comparison, Outcome, and Timeframe) framework. Population: adults aged ≥18 years with confirmed T2DM, resident in India, in any healthcare or community setting. Intervention (exposure): none evaluated; for the second objective, prespecified clinical and demographic factors (age, diabetes duration, glycaemic control, hypertension, neuropathy, physical activity and body composition) were treated as exposures. Comparison: not applicable for prevalence; for associated factors, adults with T2DM without sarcopenia or without the factor of interest. Outcome: prevalence of AWGS-defined phenotypes (primary: confirmed sarcopenia by AWGS 2019) and ORs for associated factors. Timeframe: observational studies from database inception to 14 January 2026.

Eligibility criteria

Studies were eligible if they (1) enrolled adults aged ≥18 years with confirmed T2DM; (2) were conducted in India; (3) diagnosed sarcopenia by AWGS criteria (AWGS 2019 primary; AWGS 2014 for sensitivity analysis); (4) reported prevalence for at least one AWGS phenotype or extractable effect estimates for prespecified factors; and (5) used a nonrandomised observational design. Studies of type 1 diabetes, secondary diabetes, pre-diabetes or gestational diabetes, or using non-AWGS frameworks exclusively [eg, European Working Group on Sarcopenia in Older People (EWGSOP), Strength, Assistance in walking, Rise from a chair, Climb stairs, Falls (SARC-F) questionnaire alone], were excluded, as were randomised trials, qualitative studies, editorials, case reports and letters without primary data.

Outcomes and estimate selection

The primary outcome was the prevalence of confirmed sarcopenia defined by AWGS 2019: low appendicular skeletal muscle mass index (DXA<7.0 kg/m² in men and <5.4 kg/m² in women; BIA<7.0 and <5.7 kg/m²) with low handgrip strength (<28 kg in men; <18 kg in women) or low physical performance (six-metre gait speed<1.0 m/s, Short Physical Performance Battery≤9, or five-times chair-stand≥12 s).⁶ Secondary outcomes were possible sarcopenia, severe sarcopenia and sarcopenic obesity as operationalised by each study, and associations with prespecified factors. Sarcopenia had to be ascertained by in-study instrumental assessment; self-reported or medical-record diagnoses were not acceptable.

Where a study reported multiple eligible estimates, one was selected using prespecified rules: whole-cohort estimates were preferred over subgroup estimates (eg, the total-population rather than the cardiovascular disease subgroup estimate in Wasir et al¹⁶); AWGS 2019 was preferred over AWGS 2014; and adjusted ORs were preferred over unadjusted, with the sarcopenic obesity definition applied by each study recorded at extraction.

Information sources and search strategy

A comprehensive search was conducted in PubMed (MEDLINE), Embase, Scopus and Web of Science from database inception to 14 January 2026. Controlled vocabulary and free-text terms covered sarcopenia, muscle wasting, handgrip strength, skeletal muscle mass, T2DM, India and 13 Indian geographic terms. Full line-by-line strategies—platform searched, date run, coverage dates, limits applied and records retrieved per database—are provided in online supplemental table S1, formatted in accordance with PRISMA-S.²³ No language, date or publication-type restrictions and no search filters were applied, and no previously published strategies were reused. The strategy was developed by the review team and was not independently peer reviewed. Reference lists of all included studies were hand-searched and forward citation searching was performed in Google Scholar on 14 January 2026; no study registries, grey-literature sources or organisational websites were searched, and no study authors were contacted. The search was re-run in all four databases on 9 August 2026 using identical strategies.

Study selection and data extraction

Records were deduplicated using Zotero and Rayyan. Titles, abstracts and full texts were screened independently by two reviewers (SR, RS), with disagreements resolved by consensus (SS). Extracted items included author, year, state and region, setting, sample size, age, sex, body mass index, HbA1c, diabetes duration, muscle mass modality (DXA or BIA), AWGS version, phenotype prevalences and effect estimates (ORs with 95% CIs) for prespecified factors. Unadjusted and adjusted ORs were extracted separately with model type recorded.

Risk of bias assessment

Methodological quality was assessed using the Joanna Briggs Institute Critical Appraisal Checklist for Studies Reporting Prevalence Data²⁴ (nine items: sample frame appropriateness, sampling method, sample size adequacy, subject and setting description, data coverage, validity of diagnostic methods, reliability of measurement, appropriateness of statistical analysis, response rate management). Overall risk was classified as low (≥7 items ‘Yes’), unclear (5–6) or high (≤4 or critical domain concerns). No studies were excluded on bias grounds.

Patient and public involvement

Patients or the public were not involved in the design, conduct, reporting, or dissemination plans of this research.

Statistical analysis

All analyses were performed in R V.4.5.1 using the meta (v8.2-0) and metafor (v4.8-0) packages. Individual study proportions were logit-transformed; between-study variance (τ²) was estimated by DerSimonian-Laird. CIs for pooled estimates used the Hartung-Knapp (HK) adjustment,²⁵ appropriate for k<10 studies. PIs, computed as pooled estimate±t(k−1)√(τ²+SE²) on the logit scale, quantified the expected range in a hypothetical new Indian T2DM study. Statistical heterogeneity was quantified by I² (<25% low; 25–75% moderate; >75% high) and τ²; Cochran’s Q was evaluated at α=0.10.

The primary analysis was restricted to six AWGS 2019 studies (k=6), with a sensitivity analysis incorporating the single AWGS 2014 study (Yogesh et al 2024b²⁰; k=7). Subgroup analyses were prespecified by phenotype and diagnostic instrument (DXA vs BIA), supplemented by a structured narrative comparison of study-level characteristics. Associations with prespecified factors were synthesised in a tiered framework: Tier A (adjusted odds ratios (AORs) reported by≥2 studies), Tier B (single-study AORs) and Tier C (unadjusted ORs, supplementary). Because only two studies contributed to each Tier A factor, pooling with the HK adjustment yields extremely wide intervals reflecting variance inflation at k=2; Tier A evidence is therefore synthesised narratively, in line with Synthesis Without Meta-analysis guidance,²⁶ with exploratory pooled estimates in online supplemental table S5 and online supplemental figure S10 for transparency only. Leave-one-out (LOO) analyses were performed for subgroups with k≥3. A contour-enhanced funnel plot was generated for visual inspection of publication bias; formal Egger’s regression test was not performed given k<10.²⁷ Certainty of evidence was assessed using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) framework.

Results

Study selection

The systematic search identified 271 records: PubMed (n=31), Embase (n=137), Web of Science (n=49) and Scopus (n=54). After removal of 113 duplicates, 158 unique citations underwent title and abstract screening, of which 141 were excluded as irrelevant. 17 full-text articles were assessed for eligibility and 10 were excluded, leaving seven studies in the final synthesis (figure 1, which details the reasons for exclusion).^14–20 Full citations for the 10 reports excluded at full-text assessment, with the reason for each exclusion, are provided in online supplemental table S6.

Figure 1. PRISMA 2020 flow diagram illustrating the systematic literature search and study selection process.

The updated search (9 August 2026) identified one further eligible report: Salman et al,²⁸ a DXA-based AWGS 2019 study of 151 adults with T2DM in Uttar Pradesh reporting 44% sarcopenic obesity. Because it enrolled exclusively obese participants, its denominator is not comparable with the unselected cohorts pooled here; consistent with our prespecified preference for whole-cohort over subgroup estimates it was not pooled, and is considered narratively. No other new eligible studies were identified.

Study characteristics

All seven studies employed a cross-sectional design (2023–2025), collectively enrolling 1359 adults with T2DM (table 1). Sample sizes ranged from 60 to 404, mean age from 46.2 to 65.2 years and male participants from 45% to 80%. Studies came from North India (Delhi,¹⁴ Haryana,¹⁶ Chandigarh¹⁸), West India (Gujarat,^{19 20} Maharashtra¹⁷) and East India (Odisha¹⁵); none were from South India. Six were tertiary care based and one camp-based.²⁰ DXA was used in two studies (n=388) and BIA in five (n=971). Six applied AWGS 2019 criteria; Yogesh et al 2024b²⁰ applied AWGS 2014; for confirmed sarcopenia it was included in the sensitivity analysis only.

Table 1. Characteristics of included studies.

Study	N	Age (years)	Male (%)	Region (setting)	Instrument	Confirmed sarcopenia (%)	HbA1c (%)	DM duration
Shahi 202314	229	46.2±7.4	45	North (Delhi) Tertiary	DXA	18.8% (43)	7.57±1.29	Median 3 years (IQR 2–6)
Sravya 202315	159	57.4±6.0	52	East (Odisha) Tertiary	DXA	22.0% (35)	8.4±1.9	8.2±5.87 years
Wasir 2025‡16	60	57.5±9.6	80	North (Haryana) Tertiary	BIA	23.3% (14)	NR	11.6±7.5 years
Thorat 202517	157	53.4±5.2	46	West (Maharashtra) Tertiary	BIA	27.4% (43)	8.54±1.55	NR
Lamba 202518	100	≥65	52	North (Chandigarh) Tertiary	BIA	30.0% (30)	NR	NR
Yogesh 2024a19	404	55.0±13.5	54.5	West (Gujarat) Tertiary	BIA	45.3% (183)	NR*	NR*
Yogesh 2024b†	250	65.2±5.8	60.4	West (Gujarat) Camp	BIA	60.4% (151)	7.9±1.1	NR

^*HbA1c not collected per study limitations; DM duration reported in categories only.

^†AWGS 2014 criteria applied; confirmed sarcopenia included in sensitivity analysis only.

^‡Wasir et al: whole-cohort values (n=60); the study enrolled 30 participants with and 30 without cardiovascular disease, with arm-specific diabetes duration medians of 8.5 and 10.5 years, respectively.

BIA, bioelectrical impedance analysis; DM, diabetes mellitus; DXA, dual-energy X-ray absorptiometry; HbA1c, glycated haemoglobin; IQR, interquartile range; NR, not reported.

Risk of bias

Four studies were rated low overall risk of bias (Shahi et al¹⁴, Sravya et al¹⁵, Yogesh et al 2024a¹⁹, Yogesh et al 2024b²⁰). Two studies (Thorat et al, Lamba et al) were rated unclear owing to unreported sampling methods and response rates. Wasir et al was rated high risk owing to a small pilot sample (n=60) without a formal power calculation and deliberate enrichment for cardiovascular disease (half the cohort). Bias assessments are presented in online supplemental table S3 and online supplemental figures S2-S3.

Confirmed sarcopenia prevalence

In the primary analysis restricted to AWGS 2019 studies (k=6, n=1109), the pooled confirmed sarcopenia prevalence was 27.4% (95% HK-CI 18.9% to 37.9%; 95% PI: 6.9% to 65.6%; I²=91.7%; τ²=0.34; Cochran’s Q p<0.001) (figure 2, table 2; online supplemental figure S4). The sensitivity analysis incorporating the AWGS 2014 study (Yogesh et al 2024b²⁰ ; k=7, n=1359) yielded a modestly higher pooled prevalence of 31.6% (95% HK-CI 19.8% to 46.4%; 95% PI: 6.5% to 75.5%; I²=95.1%; online supplemental figure S5). A generalised linear mixed model sensitivity analysis produced a substantively identical estimate (31.6%; 95% CI 20.2% to 45.6%).

Figure 2. Forest plots of pooled sarcopenia phenotype prevalences (confirmed sarcopenia AWGS 2019, k=6, with sensitivity k=7; possible sarcopenia, k=3; severe sarcopenia, k=5; sarcopenic obesity, k=4) with 95% Hartung–Knapp CIs and prediction intervals.

Table 2. Pooled prevalence of sarcopenia phenotypes.

Analysis	k	N	Pooled estimate	95% HK-CI	95% PI	I² (%)	τ²
Sarcopenia phenotype prevalences
Confirmed sarcopenia—AWGS 2019	6	1109	27.4%	18.9% to 37.9%	6.9%–65.6%	91.7	0.34
Confirmed sarcopenia—sensitivity (AWGS 2019+2014)	7	1359	31.6%	19.8% to 46.4%	6.5%–75.5%	95.1	0.52
Possible sarcopenia	3	693	65.6%	6.7% to 98.1%	0.1%–100%	98.6	2.13
Severe sarcopenia	5	705	11.0%	6.1% to 19.0%	3.2%–31.8%	68.9	0.18
Sarcopenic obesity	4	943	25.2%	10.2% to 50.0%	3.8%–74.3%	93.2	0.36
Subgroup: DXA (AWGS 2019)	2	388	20.2%	6.7% to 47.0%	—	0	0
Subgroup: BIA (AWGS 2019)	4	721	31.7%	18.8% to 48.4%	—	87.7	0.23

Phenotype denominators (N) differ across analyses because not all studies reported every phenotype (see text). Exploratory pooled adjusted odds ratios for factors associated with sarcopenia are presented in online supplemental table S5 and figure S10.

AWGS, Asian Working Group for Sarcopenia; BIA, bioelectrical impedance analysis; DXA, dual-energy X-ray absorptiometry; HK-CI, Hartung-Knapp confidence interval; I², heterogeneity statistic; PI, prediction interval; τ², between-study variance (rounded to two decimal places).

Prespecified subgroup analysis by diagnostic instrument (AWGS 2019 studies only) gave a pooled prevalence of 20.2% (95% HK-CI 6.7% to 47.0%; I²=0%) in the DXA subgroup (k=2) and 31.7% (95% HK-CI 18.8% to 48.4%; I²=87.7%) in the BIA subgroup (k=4); the test for subgroup differences was significant (Qbetween=6.41, p=0.011; online supplemental figure S9). Instrument modality therefore accounted for a substantial share of the heterogeneity, with none remaining within the DXA subgroup and marked heterogeneity persisting within the BIA subgroup, largely attributable to Yogesh et al 2024a¹⁹ (see Sensitivity analyses).

Sarcopenia phenotype spectrum

Pooled possible sarcopenia prevalence (k=3; n=693) was 65.6% (95% HK-CI 6.7% to 98.1%; 95% PI: 0.1% to 100%; I²=98.6%), reflecting extreme heterogeneity from markedly different screening operationalisations; with a prediction interval spanning almost the entire possible range, this value is a descriptive summary of three discordant estimates rather than an interpretable prevalence (online supplemental figure S6). Pooled severe sarcopenia prevalence (k=5; n=705) was 11.0% (95% HK-CI 6.1% to 19.0%; 95% PI: 3.2% to 31.8%; I²=68.9%; online supplemental figure S7) and sarcopenic obesity prevalence (k=4; n=943) was 25.2% (95% HK-CI 10.2% to 50.0%; 95% PI: 3.8% to 74.3%; I²=93.2%; online supplemental figure S8). Denominators differ because not every study reported every phenotype: possible sarcopenia was unavailable from three studies^{15 17 18} and not assessable under AWGS 2014,²⁰ severe sarcopenia from Yogesh et al 2024a¹⁹ and 2024b,²⁰ and sarcopenic obesity from three studies.^{15 17 18} All phenotype results are presented in table 2 and figure 2.

The certainty of evidence was very low for all prevalence outcomes (GRADE), primarily driven by risk of bias concerns, substantial inconsistency and imprecision (online supplemental table S4).

Sensitivity analyses

LOO analysis for the primary AWGS 2019 estimate (k=6) gave pooled prevalences from 23.9% (omitting Yogesh et al 2024a) to 29.5% (omitting Shahi et al).¹⁴ Omitting Yogesh et al 2024a reduced I² from 91.7% to 39.9%, identifying it as the primary heterogeneity driver, likely reflecting its large sample and highest individual prevalence (45.3%). In the k=7 analysis, pooled prevalence ranged from 27.4% to 34.3%.

Narrative exploration of heterogeneity

With seven studies, formal meta-regression was not feasible; other sources of heterogeneity were explored through structured narrative comparison. Three patterns emerged. First, age structure: cohorts of older adults reported the highest prevalences (Lamba et al,¹⁸ all≥65 years: 30.0%; Yogesh et al 2024b,²⁰ mean 65.2 years: 60.4%) and the youngest cohort the lowest (Shahi et al,¹⁴ mean 46.2 years: 18.8%). Second, setting and case mix: the sole camp-based study²⁰ reported the highest prevalence overall, and Wasir et al¹⁶ deliberately over-represented cardiovascular disease (half the cohort). Third, AWGS version: excluding the AWGS 2014 study returned the sensitivity estimate to the primary value. Geographic region showed no consistent gradient. These characteristics are confounded with one another and with instrument modality, and are hypothesis-generating.

Factors associated with sarcopenia

Tier A (adjusted estimates reported by two studies): both were rated at low risk of bias but differed substantially in population and methods. Sravya et al¹⁵—a tertiary-care, DXA-based, AWGS 2019 study of 159 middle-aged outpatients in Odisha—reported an AOR for hypertension of 8.74 (95% CI 1.91 to 39.92); Yogesh et al 2024b²⁰—a camp-based, BIA-based, AWGS 2014 study of 250 older adults in Gujarat—reported 2.3 (95% CI 1.5 to 3.8), from models retaining different covariate sets (online supplemental table S2). Both indicated a positive association but each carried wide uncertainty; the magnitude and generalisability remain unclear.

Diabetes duration≥10 years likewise showed a positive association in both studies, reaching significance in one (Yogesh et al 2024b²⁰: AOR 4.1, 95% CI 2.9 to 8.9; Sravya et al¹⁵: unadjusted OR 1.71, 95% CI 0.80 to 3.66). Exploratory pooling of these k=2 comparisons produced uninterpretably wide Hartung–Knapp intervals (hypertension: 3.66, 95% HK-CI 0.001 to 11 748; duration, combining one adjusted and one unadjusted estimate: 2.75, 95% HK-CI 0.011 to 697), reflecting variance inflation at k=2 rather than evidence against an association; these are reported for transparency only (online supplemental table S5 and online supplemental figure S10).

Tier B (single-study adjusted estimates): diabetic neuropathy (AOR 5.57, 95% CI 1.26 to 24.66; Sravya et al¹⁵), age≥65 years (2.3, 1.2 to 4.5), uncontrolled HbA1c≥8% (2.8, 1.2 to 6.1) and low physical activity (2.8, 1.3 to 6.2; all Yogesh et al 2024b²⁰) were each significant in a single study only, and none has been replicated in another Indian T2DM cohort. Higher BMI was consistently inversely associated with sarcopenia (AOR 0.20, 0.09 to 0.44, Shahi et al¹⁴; 0.019, 0.001 to 0.248, Sravya et al¹⁵).

Tier C (unadjusted associations, online supplemental table S2): Yogesh et al 2024a¹⁹ reported crude ORs for age≥60 years (2.61, 1.7 to 3.9), diabetes duration≥7 years (7.5, 3.65 to 15.4) and body fat percentage-defined obesity (2.2, 1.47 to 3.4). These are susceptible to confounding and were not pooled or synthesised with the adjusted estimates.

Publication bias

A contour-enhanced funnel plot for confirmed sarcopenia (k=7) is presented in online supplemental figure S1. Visual inspection showed mild asymmetry, with a possible deficit of smaller studies reporting lower prevalence; formal tests were not performed (k<10),²⁷ so publication bias cannot be assessed definitively.

Discussion

Principal findings

This is the first India-specific, AWGS 2019-based synthesis of sarcopenia prevalence in adults with T2DM. Confirmed sarcopenia affected approximately one in four Indian adults with T2DM (27.4%, 95% HK-CI 18.9% to 37.9%; k=6; n=1109), numerically higher than the approximately 18% reported across predominantly Asian and European T2DM cohorts,¹⁰ the 17% pooled across Asia by Wu et al¹² and the 23% pooled across Asian AWGS 2019 studies by Yogesh et al.¹³ The prediction interval nevertheless extends from 6.9% to 65.6% and certainty was very low for every outcome, so this ordering signals a plausibly higher Indian burden rather than an established difference.

Strengths and limitations

Strengths include prospective PROSPERO registration, adherence to PRISMA 2020, PRISMA-S and MOOSE guidance, dual independent source-verified extraction, and the tiered synthesis framework. Limitations include cross-sectional designs precluding causal inference, recruitment confined to tertiary or camp settings (likely overestimating community prevalence), marked heterogeneity, few studies limiting subgroup power, non-poolable sex-disaggregated data, a DXA subgroup of k=2, no South Indian studies, and very low certainty throughout. Publication bias cannot be quantified, and with only two studies per shared factor no factor-level certainty assessment was possible.

Comparison with existing literature

Several factors may explain the higher Indian prevalence. The ‘thin-fat’ phenotype is a substrate particularly vulnerable to sarcopenia,^{7 8} and nationally representative data indicate that 4%–26% of Indians, depending on age group and residence, are at risk of inadequate-quality protein intake against a recommended 1.0–1.2 g/kg/day.^{29 30} Insulin resistance further disrupts anabolic signalling and accelerates proteolysis.³ Recruitment confined to tertiary care or camp settings selectively captures established or complex disease, which may itself inflate prevalence relative to community-representative samples.

The difference in prevalence by measurement modality (20.2% DXA vs 31.7% BIA; Qbetween=6.41, p=0.011) is methodologically important. BIA is susceptible to the extracellular fluid shifts inherent to T2DM, including subclinical oedema and glycaemic dysregulation-mediated hydration changes, which can distort muscle mass estimates,^{31 32} whereas DXA is not subject to these artefacts and remains the reference standard.³³ Future Indian studies should prioritise DXA standardisation.

Our pooled Indian prevalence (27.4%) sits above the pan-Asian estimates of Wu et al (17%, 43 studies) and Yogesh et al (23%, 39 studies),^{12 13} but the gap narrows once measurement modality is considered. Our DXA subgroup (k=2, n=388) yielded 20.2% with no residual heterogeneity, and Wu et al report an India subgroup of 20% (95% CI 16% to 24%; I²=0%) from the same two DXA-based cohorts,^{14 15} an independent replication of the most measurement-robust component of our analysis. Conversely, the five studies forming the elevated South-East Asia subgroup of Yogesh et al (37.5%) were all conducted in India, and that estimate brackets our BIA subgroup (31.7%). The phenotypic spectrum likewise agrees closely with Yogesh et al¹³: possible sarcopenia 65.6% versus 61%, severe sarcopenia 11.0% versus 12.1% and confirmed sarcopenia 27.4% versus 23%.

Two divergences deserve emphasis. First, we found DXA-based prevalence significantly lower than BIA-based, whereas both pan-Asian reviews found DXA numerically higher (Wu et al 19% vs 16%; Yogesh et al 29.9% vs 19.5%, p=0.71), neither significantly; with only two DXA studies, instrument modality in our data is confounded with age structure, setting and AWGS version, so this subgroup finding should not be over-interpreted. Second, although Yogesh et al pooled adjusted associations for hypertension (3.65, 95% CI 1.06 to 12.65) and diabetes duration (1.35, 95% CI 1.05 to 2.13) in the same direction as our Tier A findings, Wu et al found neither significant in a larger evidence base, so these associations remain unresolved even at the pan-Asian level.

Heterogeneity and its implications

The very high heterogeneity of the primary estimate (I²=91.7%) is central to interpretation, not merely a limitation. The pooled 27.4% is an average across markedly different clinical contexts rather than the value expected in any particular setting: the prediction interval indicates that a new Indian T2DM cohort could plausibly observe anywhere between approximately 7% and 66%. Measurement modality—the only source formally testable by subgroup analysis—explained a substantial share, and within the BIA subgroup leave-one-out analysis identified Yogesh et al 2024a¹⁹ as the dominant driver. Practically, the DXA subgroup estimate (20.2%) may be the more conservative, measurement-robust benchmark, and citing 27.4% without its confidence and PIs would overstate its precision.

Factors associated with sarcopenia: interpretation and reliability

The evidence on associated factors is sparse, resting on two adjusted analyses in different populations with different instruments and AWGS versions. The direction for hypertension is biologically plausible: the conditions share candidate pathways including chronic low-grade inflammation, renin–angiotensin activation, microvascular dysfunction and physical inactivity, and hyperglycaemia is an established correlate of muscle loss.^{34 35} Reliability is nonetheless limited: intervals were wide, covariate sets differed, Tier B associations are unreplicated, and cross-sectional design cannot establish temporality. In the study from our updated search, hypertension was not significantly associated with sarcopenic obesity.²⁸ These factors are best interpreted as cross-sectional correlates rather than causal determinants.

Meaning of the study and implications

Severe sarcopenia (11.0%) was the most internally consistent phenotype, indicating that roughly 1 in 9 patients meets the highest-severity AWGS 2019 classification. Even so, these findings are not strong enough to support a recommendation for routine sarcopenia screening in Indian T2DM care. Certainty is very low, all estimates derive from seven small cross-sectional studies in tertiary care or camp settings, and prevalence in a new Indian cohort could plausibly lie between approximately 7% and 66%. A screening policy cannot be calibrated against a range that wide, and no study in this population has shown that screening or subsequent intervention improves patient-important outcomes. What the evidence supports is greater clinical awareness: low muscle mass and strength are likely to be common and currently seldom looked for, and assessment may reasonably be considered in patients with longer diabetes duration, hypertension, older age or lower body mass index. Whether it should become routine, and at what threshold, requires adequately powered studies. Large-scale, community-based, DXA-standardised prospective studies with sex-stratified designs, including from South India, remain the priority.

Data availability statement

Data are available upon reasonable request. All data relevant to the study are included in the article or uploaded as supplementary information.