Buffalo Hump, Neck Circumference, and Stroke Risk: A Narrative Review from a Surface Landmark to an Untested Intervention Target

1. Introduction

Buffalo hump is a common soft-tissue prominence at the cervicothoracic junction and is also referred to in the clinical literature as a dorsocervical fat pad or dorsocervical fibro-lipodystrophy. Its traditional conceptualization has long been confined to fat accumulation and cosmetic concerns and has not yet been systematically examined within a stroke risk framework. Meanwhile, neck circumference, as a simple anthropometric measure, has accumulated relatively consistent evidence in cardiometabolic epidemiology. The Framingham Heart Study demonstrated in 3,307 participants that neck circumference was continuously and dose-dependently associated with multiple metabolic measures, including systolic blood pressure, diastolic blood pressure, triglycerides, fasting glucose, insulin, insulin resistance, and high-density lipoprotein cholesterol; these associations remained significant even after full adjustment for visceral fat and body mass index (BMI) (Preis et al., 2010). This single-cohort finding was subsequently supported by integrated evidence from systematic reviews and meta-analyses: the associations of neck circumference with blood pressure, blood lipids, blood glucose, and insulin resistance were directionally consistent across studies (Ataie-Jafari et al., 2018), and the association of neck circumference with metabolic syndrome and its components was also systematically integrated (Namazi et al., 2018). However, the endpoints of these integrated analyses remain metabolic risk factors or metabolic syndrome rather than hard endpoints such as stroke. This association spans multiple dimensions, including blood pressure, glucose metabolism, lipid metabolism, and insulin sensitivity, suggesting that upper-body subcutaneous fat may carry systemic pathophysiological information independent of overall obesity. This methodological positioning provides the necessary logical starting point for linking neck circumference to hard endpoints such as stroke.

Li G et al. (2024), in a multicenter prospective cohort of 18,796 participants in northeastern China, found that neck circumference was associated with an increased risk of cardiovascular mortality, with predictive value independent of other obesity indices. Cao et al. (2022), in a community-based longitudinal follow-up study, reported that higher neck circumference was associated with an increased risk of new-onset carotid plaque and carotid intima-media thickness (CIMT) thickening during follow-up. Between mortality endpoints and subclinical endpoints, Yang GR et al. (2019), using the 8-year prospective Beijing Community Diabetes Study 20 (BCDS-20) cohort of patients with type 2 diabetes, provided evidence at the level of incident hard endpoints: among 3,009 patients with type 2 diabetes, high neck circumference was independently associated with an increased risk of a composite cardiovascular endpoint that included stroke, with an adjusted hazard ratio (HR) of approximately 2.3. This estimate should not be interpreted as a stroke-specific hazard ratio. These findings establish an important premise: an increase in neck circumference is a marker of stroke risk. However, for a marker to be translated into a target for prevention strategies, two further questions need to be answered. First, does enlargement of the buffalo hump lead to an increase in neck circumference, thereby making the buffalo hump a modifiable starting point for intervention? Second, can interventions such as surgery, lifestyle modification, physical therapy, or exercise training reduce the volume of the buffalo hump or neck fat, thereby lowering neck circumference and ultimately leading to a reduction in stroke events? The goal of this article is not to argue that the answers to these two questions have already been established but rather to review the existing evidence surrounding these two gaps and to identify their structural gaps.

To make the argument explicit, this review evaluates the evidence along four sequential links: (1) buffalo hump to neck circumference; (2) neck circumference to stroke risk; (3) intervention to a clinically meaningful reduction in neck circumference; and (4) reduction in neck circumference to reduction in stroke events. Link 1 is currently supported mainly by anatomical inference rather than same-population measurement. Link 2 is supported by observational and longitudinal evidence, with important heterogeneity across populations and endpoints. Link 3 has preliminary technical evidence but lacks systematic reporting of neck circumference as an outcome in buffalo hump populations. Link 4 remains untested. This four-link framework is used throughout the review to distinguish measured associations from inferred or untested transitions.

This article is a narrative review. Literature searches were conducted in PubMed, Embase, the Cochrane Library, and the China National Knowledge Infrastructure, with a search cutoff of March 2026. The English search strategy combined the following terms: (“buffalo hump” OR “dorsocervical fat” OR “neck fat pad” OR “fibro-lipodystrophy”) AND (“neck circumference” OR “neck adiposity”) AND (“stroke” OR “cerebral infarction” OR “carotid atherosclerosis” OR “cardiovascular events”). It was supplemented by (“physical therapy” OR “exercise” OR “posture correction” OR “upper crossed syndrome”) AND (“neck circumference” OR “cervical fat”) for intervention studies. The Chinese search strategy combined terms such as “buffalo hump,” “dorsocervical fat pad,” “neck circumference,” “stroke,” “physical therapy,” and “upper crossed syndrome.” Inclusion criteria were population studies, intervention studies, mechanistic studies, or systematic reviews/meta-analyses that reported neck circumference, buffalo hump/dorsocervical fat, carotid imaging, or stroke/cardiovascular endpoints; prospective cohorts, randomized controlled trials, and systematic reviews were prioritized. Exclusion criteria were case reports alone (unless they provided unique mechanistic clues), narrative reviews without quantitative outcomes, and duplicate publications. Search records. Searches were run in the four databases using the strategies described above. Database export logs with retrieval dates and search strings were archived; because this was a narrative review, records retrieved per database, duplicate counts, and full-text exclusion counts were not systematically tabulated, and no PRISMA flowchart was generated. Screening was performed by a single author without independent dual screening. A risk of selection bias therefore exists, and this article does not claim exhaustive retrieval according to systematic review standards. After screening, 42 formally published articles were included in this review. The search strings and available records are available from the authors upon request. Because this is a narrative review, we did not perform formal GRADE assessment, risk-of-bias scoring, or article-by-article evidence labeling. Instead, we characterized the strength of evidence for each of the four links descriptively, considering consistency, directness, study design, precision, effect size, and adequacy of adjustment for adiposity. Link 1 was characterized as inferential because no same-population measurement of buffalo hump and neck circumference was identified. Link 2 was characterized as supportive but inconsistent: large prospective cohorts with substantial effect sizes support an association between neck circumference and stroke risk, but negative results, inconsistent sex modification, and the predominance of composite cardiovascular endpoints limit the strength of inference. Link 3 was characterized as preliminary because technical fat reduction has not been systematically linked to neck circumference in buffalo hump populations. Link 4 was characterized as absent because no direct interventional or longitudinal evidence relates neck circumference reduction to stroke events. These characterizations are intended to make the narrative synthesis transparent rather than to imply a systematic review-level certainty rating.

2. Morphologic Heterogeneity and Metabolic Comorbidity Features of Buffalo Hump

Equating buffalo hump with localized fat accumulation is directly challenged by large-sample imaging classification studies. Li FW et al. (2026) conducted a clinical and magnetic resonance imaging (MRI) classification study of 383 patients with buffalo hump. The study showed that non-HIV-related buffalo hump can be morphologically divided into four types: type I (dorsocervical fat pad), 77.57%; type II (fat accumulation), 11.23%; type III (lipoma), 10.18%; and type IV (mixed type), 1.04%. Only 1 patient in this group was HIV-positive. The vast majority of buffalo humps have an identifiable focal structure, and the extent to which they can be debulked as an intervention target may differ by subtype. Wang et al. (2022), in an MRI anatomical study and clinical case series, further localized the dorsocervical fat pad to the C3–T3 region. They reported mean measurements of approximately 114.47, 89.24, and 23.46 mm for its length, width, and thickness, respectively, and noted that the severity of the dorsocervical prominence was only weakly correlated with obesity. This finding is consistent with the direction of the fat compartmentalization research by Torriani et al. (2014), suggesting that the dorsocervical fat pad is not a simple linear function of overall obesity but has relatively independent structural and metabolic significance. At the same time, that study focused on patient satisfaction and local cosmetic improvement and did not systematically report neck circumference as an outcome variable (Wang et al., 2022); this itself constitutes a concrete example of the misalignment between intervention outcomes and stroke risk concerns discussed in Section 4.

The clinical association between buffalo hump and metabolic burden has been further characterized in non-HIV populations. Lee et al. (2025) reviewed 12 HIV-negative patients with buffalo hump. Mean BMI was 30.15 ± 4.59 kg/m²; the prevalence of hypertension and diabetes was 66.7% each, and the prevalence of hyperlipidemia was 75%. Of the 12 patients, 3 were newly diagnosed at presentation with metabolic diseases such as diabetes, hyperlipidemia, or fatty liver. This suggests that the appearance of buffalo hump may serve as a window into previously unrecognized metabolic abnormalities. In a more specific population, Cheung et al. (2007) studied 123 patients with biopsy-confirmed nonalcoholic fatty liver disease (NAFLD). They found that the prevalence of dorsocervical lipohypertrophy (DCL) was 28.5%. DCL was significantly associated with the severity of all liver histologic parameters except simple steatosis and was the single largest contributor to variability in histologic severity. It should be noted that the assessment of DCL in that study relied on visual inspection rather than imaging or volumetric quantification, and exposure measurement precision was therefore limited. Consequently, its evidentiary value lies mainly in reinforcing the directionality of co-occurrence of dorsocervical fat and metabolic disease severity rather than in providing quantitative exposure data suitable for causal inference.

Imaging studies have further advanced this association to the level of fat compartmentalization. Torriani et al. (2014), using positron emission tomography/computed tomography, divided neck fat into three compartments—posterior (NATpost), subcutaneous (NATsc), and perivertebral (NATperivert)—and found that NATpost and NATsc showed the most consistent associations with cardiovascular risk factors. After adjustment for age and BMI, each 1-standard-deviation (SD) increase in these two compartments was still associated with a prevalence ratio for metabolic syndrome of approximately 1.5. The dorsocervical region where the buffalo hump is located is precisely the core distribution area of NATpost, suggesting that abnormal expansion of fat in this region may have local pathogenic effects beyond overall obesity.

From an anatomical standpoint, the core region of the buffalo hump lies near the seventh cervical vertebra (C7); the standard measurement site for neck circumference extends from the upper border of C7 posteriorly to just below the laryngeal prominence anteriorly, and the anatomical overlap between the two is clear. However, no population study in the existing literature has directly used buffalo hump as an exposure variable and systematically reported its quantitative relationship with neck circumference (Li FW et al., 2026). Lee et al. (2025) also explicitly acknowledged in the limitations of their non-HIV buffalo hump cohort that the retrospective design precluded systematic collection of neck circumference data and quantitative information on fat volume. Therefore, Link 1 of the four-link framework—from buffalo hump to increased neck circumference—is currently an anatomical inference rather than a same-population measured association. This distinction is important because the epidemiologic evidence for neck circumference cannot be automatically transferred to buffalo hump populations without this prerequisite validation.

In addition, clinical observations show that buffalo hump frequently co-occurs with forward head posture. The two commonly appear together in the postural pattern of upper crossed syndrome, manifesting as a combination of rounded shoulders, thoracic kyphosis, forward head displacement, and winged scapulae. It should be made clear that frequent co-occurrence is not equivalent to complete coexistence: the existing literature still lacks systematic measurements of the prevalence of forward head posture specifically in populations with buffalo hump, as well as postural data stratified by buffalo hump subtype. Therefore, the relationship between buffalo hump and forward head posture should currently be described as a common combination in clinical observation rather than as a universally applicable quantitative rule or an established intermediate link.

3. Epidemiologic Evidence Linking Neck Circumference to Stroke Risk: An Intermediate Target for Prevention Strategies

Before discussing whether intervention can prevent stroke by reducing neck circumference, it is first necessary to establish whether the association between neck circumference and stroke risk is robust and what its boundaries are. This section therefore addresses Link 2 of the four-link framework: from neck circumference to stroke risk.

3.1. Prospective Cohorts: Neck Circumference, Cardiovascular Mortality, and Hard Endpoint Events

Li G et al. (2024) conducted an analysis based on a multicenter prospective cohort in northeastern China. Their study provided evidence with relatively clear temporality for the relationship between neck circumference and stroke mortality. The study included 18,796 participants. Multivariable Cox regression showed that, compared with the lowest quartile, the highest quartile of neck circumference was associated with 1.83 times the risk (95% CI, 1.29–2.61) of overall cardiovascular mortality and 2.40 times the risk (95% CI, 1.45–4.00) of coronary heart disease mortality. Each 1-SD increase in neck circumference was associated with approximately 1.21 to 1.25 times the risk of overall cardiovascular mortality, stroke mortality, and coronary heart disease mortality. In that study, neck circumference was significantly associated with the risk of ischemic stroke mortality, whereas no corresponding association was observed with the risk of hemorrhagic stroke mortality. After neck circumference was added to an existing prediction model, the correct classification rate for overall cardiovascular mortality improved by 12.1% (Li G et al., 2024).

Beyond mortality endpoints, BCDS-20 provided prospective multicenter evidence for the longitudinal association between neck circumference and the incidence of cardiovascular events, including stroke. The study included 3,009 patients with type 2 diabetes from the Beijing community, with 8 years of follow-up, and identified 211 cardiovascular events; the endpoint definition explicitly included stroke, transient ischemic attack (TIA), and hemorrhagic stroke. After stepwise adjustment for age, sex, smoking, systolic blood pressure, glycated hemoglobin, low-density lipoprotein cholesterol, diabetes duration, and medication use, high neck circumference was associated with an adjusted HR for cardiovascular events of 2.305 (95% CI, 1.535–3.460) (Yang GR et al., 2019). High neck circumference was defined as >43 cm in men and >39 cm in women, based on the upper quartile of this cohort. This adjusted model included systolic blood pressure, glycated hemoglobin, and low-density lipoprotein cholesterol, indicating that the association of neck circumference was not mediated solely by the three traditional risk factors of hypertension, hyperglycemia, and dyslipidemia. In the chain of evidence, BCDS-20 precisely fills the gap between the mortality endpoint in Li G et al. (2024) and the subclinical endpoint in Cao et al. (2022)—namely, the incidence of nonfatal hard endpoint events. Its endpoint definition explicitly includes both ischemic and hemorrhagic stroke. However, the reported HR refers to a composite cardiovascular endpoint rather than to stroke alone.

3.2. Longitudinal Follow-Up: Neck Circumference and Carotid Atherosclerosis Progression

Cao et al. (2022), in a Shanghai community-based study of 1,189 participants without carotid plaque at baseline (mean age, 59.6 ± 7.3 years), had a mean follow-up of 2.1 years. The risk of new-onset carotid plaque in the high neck circumference group (≥38.5 cm in men and ≥34.5 cm in women) was 1.55 times that in the low neck circumference group (95% CI, 1.12–2.14); among those without plaque at follow-up, the risk of new-onset CIMT thickening in the high neck circumference group was 1.49 times that in the low neck circumference group (95% CI, 1.14–1.95). Neck circumference was measured at baseline, and atherosclerotic outcomes were newly confirmed during follow-up; an increase in neck circumference preceding the progression of carotid atherosclerosis thus becomes a reasonable working description, providing longitudinal evidence for the intermediate link in the sequence from neck circumference to subclinical atherosclerosis to clinical events.

3.3. Sex Differences: Evidence with Inconsistent Directions

The literature does not support a stable direction of effect modification by sex. Li G et al. (2022), in a cross-sectional study based on a population in northeastern China, examined sex differences in the association between neck circumference and the prevalence of ischemic stroke. In women, the association between neck circumference and ischemic stroke remained robust after full adjustment for traditional risk factors, whereas in men the association was no longer statistically significant. The researchers noted that the association between cervical fat accumulation and the severity of obstructive sleep apnea (OSA) was more pronounced in women, which may partially explain the sex difference. Han and Kim (2022), using a 12-year prospective community-based cohort of 3,662 Korean adults, reported that the association between neck circumference and the incidence of cerebrovascular disease was stronger in women than in men. The cross-sectional study and the 12-year cohort thus yielded a consistent pattern in two independent populations.

However, this pattern of a more robust association in women is not consistent across cohorts. Hu T et al. (2022) conducted a 7.6-year prospective cohort of 1,435 residents aged 50–80 years in the Shanghai community. The cohort recorded 148 cardiovascular events, and the neck circumference cutoffs were consistent with those of Cao et al. (2022): ≥38.5 cm in men and ≥34.5 cm in women. The study reported sex-stratified results in the opposite direction: in men, the HR for cardiovascular events per 1-SD increase in neck circumference was 1.37 (95% CI, 1.10–1.71), and the HR for high versus low neck circumference was 1.90 (95% CI, 1.21–2.98), both significant. In women, neither reached statistical significance: the per-SD HR was 1.19 (95% CI, 0.94–1.52), and the HR for high versus low neck circumference was 1.25 (95% CI, 0.75–2.07). The composite endpoint of that study was based primarily on ischemic heart disease, and stroke was not listed individually as an endpoint in the abstract. Therefore, it cannot be directly equated with evidence on sex differences in stroke endpoints. However, its population source and cutoff definitions overlapped considerably with those of Cao et al., yet it yielded a pattern of effect modification by sex in the opposite direction. This contrast suggests that effect modification by sex is not a directionally stable phenomenon across existing East Asian cohorts. Prioritizing women as the target population for neck circumference intervention currently lacks consistent cross-cohort support. Sex differences are better regarded as an effect-modifying variable to be considered through stratification in future research rather than as an established basis for intervention prioritization.

In addition to biological effect modification, methodological variation in measurement is a competing source that cannot be ignored when explaining cross-cohort inconsistency. Sato et al. (2024) reported an interrater intraclass correlation coefficient (ICC) of 0.98 for neck circumference in older adults receiving long-term care, but the 95% limits of agreement for interrater differences still ranged from −0.9 to 1.2 cm. Hagren et al. (2023) reported test–retest ICCs of 0.85–0.94 in healthy men and women, with a standard error of measurement of approximately 1.9%–2.1%. In women with suspected OSA, Gjevre et al. (2013) reported an interrater ICC of 0.78 for neck circumference, which was below the excellent level in healthy populations but still supported its operability in clinical prediction. In addition, the systematic review and meta-analysis by Espinoza López et al. (2021) noted that measurement sites and cutoff definitions for neck circumference lack standardized criteria across studies, and the prevalence of high neck circumference in Latin America and the Caribbean can vary from 37.00% to 57.69% depending on the definition used. These data suggest that measurement error near the cutoff and differences in measurement location are sufficient to affect the classification of high versus low neck circumference. They further suggest that the inconsistency in the direction of effect modification by sex cannot be fully attributed to biological differences in fat distribution. The magnitude of the methodological contribution remains to be systematically evaluated.

3.4. Population Heterogeneity: Negative Evidence from Racially Diverse Cohorts

The evidence described above, derived mainly from East Asian populations, consistently points to a positive association, but this association is not robustly replicated in all populations. The Jackson Heart Study evaluated the relationship between neck circumference and cardiovascular outcomes in 5,290 African American participants. After adjustment for age and sex, a larger neck circumference was associated with an increased risk of heart failure hospitalization, but no independent association was observed with stroke, myocardial infarction, coronary heart disease, or all-cause mortality (Pumill et al., 2019). In a large prospective cohort on the European continent (the German National Cohort), the association of neck circumference with cardiometabolic risk factors and diseases was corroborated (Strathmann et al., 2025); however, the endpoint of that study was not stroke, and its relationship with hard-endpoint evidence for stroke cannot be directly reconciled, suggesting that the predictive value of neck circumference may be stratified by endpoint type and population background. An analysis from the Framingham Heart Study Offspring Cohort in the United States (predominantly White) showed that each 1-SD increase in neck circumference was associated with a 0.071-mm increase in internal carotid artery intima-media thickness, which remained significant after adjustment for BMI. However, during a mean follow-up of 6.1 years (147 stroke events), none of neck circumference, BMI, or waist circumference was significantly associated with incident stroke (Rosenquist et al., 2013). This negative result contrasts with the positive findings in East Asian populations, suggesting that the predictive value of neck circumference may be influenced by racial composition, endpoint type, and population background. Direct extrapolation of conclusions from East Asian cohorts to stroke prevention strategies in other populations therefore warrants caution.

3.5. Extended Evidence on Neck Circumference and Post-Stroke Prognosis

Beyond stroke incidence and mortality risk, the clinical significance of neck circumference may also extend to post-stroke prognostic assessment. The prospective study by Medeiros et al. (2011) included 89 patients with acute ischemic stroke, with 12 months of follow-up; the results showed that neck circumference was significantly larger in nonsurvivors than in survivors (P = 0.02). Because the prevalence of obesity in that cohort was only 11.2%, the researchers explicitly stated that a larger neck circumference should be regarded as a significant clinical sign associated with mortality rather than merely a proxy indicator of obesity. The prospective study by Li Y et al. (2021) included 312 patients with spontaneous intracerebral hemorrhage and found that neck circumference was an independent predictor of poor 6-month functional outcome (modified Rankin Scale score), with an odds ratio (OR) of 1.205 (95% CI, 1.075–1.350). In that cohort, neck circumference was positively associated with poor prognosis, whereas BMI, in the context of the obesity paradox, may show an opposite trend, further suggesting that neck circumference and BMI reflect fundamentally different pathophysiological information. Both studies had limited sample sizes, and their conclusions require validation in larger, multicenter stroke prognosis cohorts; however, with respect to the clinical significance of neck circumference as a surface landmark, this evidence complements the continuum from risk prediction to prognostic assessment. The evidence base spanning risk prediction and prognostic assessment is now relatively complete; however, for a marker to become a target for intervention, the question of modifiability must still be addressed—a question discussed in the next section.

4. Clinical Evidence for Interventions and Gaps in Translation to Stroke Prevention

The aforementioned epidemiologic evidence establishes neck circumference as an intermediate target for stroke risk, but whether the buffalo hump or cervical fat is modifiable and whether intervention can translate into stroke prevention require answers from intervention studies. To avoid conflating different levels of evidence, this section evaluates Link 3 and Link 4 of the four-link framework: whether an intervention can produce a clinically meaningful reduction in neck circumference (Link 3), and whether that reduction translates into fewer stroke events (Link 4). The existing literature provides preliminary evidence mainly for the technical feasibility of reducing buffalo hump or cervical fat; evidence for Link 3 is fragmentary, and Link 4 remains untested.

Regarding surgical debulking, the systematic review and meta-analysis by Escandón et al. (2026) quantitatively synthesized the safety and efficacy of surgery for dorsocervical fibro-lipodystrophy. Liposuction was the most commonly reported procedure, with a pooled recurrence rate of 10.6% and a revision rate of 15.3%; after sensitivity analysis, the recurrence and revision rates decreased to 1.4% and 1.3%, respectively. The pooled recurrence rate for excision and hybrid techniques was 0%, but the complication rate was higher. The authors concluded that liposuction is a safe and effective treatment option when performed with meticulous technique, while noting that the overall quality of evidence remains to be improved. This level of evidence provides relatively systematic support for the claim that the buffalo hump is technically debulkable, but it does not address Link 3 or Link 4. Wang et al. (2022) also reported local cosmetic improvement and patient satisfaction in their liposuction case series, but likewise did not systematically measure neck circumference as an outcome variable. This indicates that a structural misalignment persists between the outcomes assessed in the intervention literature and the concerns of stroke risk reduction.

Regarding lifestyle intervention (diet), the CENTRAL randomized controlled trial provides the highest-level evidence to date on whether non-surgical approaches can reduce cervical fat. The study included 278 participants with abdominal obesity or dyslipidemia, randomly assigned them to different dietary patterns with or without exercise intervention, and conducted 18-month MRI follow-up. Cervical subcutaneous fat decreased significantly by 13.1% at 6 months and remained 5% lower at 18 months; the 18-month reduction in cervical fat was independently associated with decreases in triglycerides and leptin (Tsaban et al., 2021). This demonstrates that lifestyle intervention can effectively reduce cervical fat depots and is accompanied by improvements in metabolic indicators. However, the target tissue of that study was cervical subcutaneous fat rather than specifically the dorsocervical fat pad where the buffalo hump is located; its outcomes were fat area and metabolic indicators rather than neck circumference or stroke endpoints. It therefore informs the technical starting point but does not address Link 3 or Link 4.

Regarding physical therapy, existing studies are mostly small-sample exploratory investigations, and within the scope of the search and inclusion for this review, no randomized controlled trial of physical therapy with neck circumference as an outcome was identified. Therefore, the effect of physical therapy on neck circumference improvement relative to control, as well as its effect size, cannot be quantitatively assessed within the current evidence framework. The strength of evidence for this pathway is insufficient to support any inference regarding stroke prevention and should be listed as a gap that future intervention research needs to address as a priority.

Regarding exercise training, the ACTIBATE randomized controlled trial provides direct evidence on whether exercise can alter neck circumference. The study included 74 young adults, randomly assigned them to a control group, a moderate-intensity exercise group, and a vigorous-intensity exercise group, and conducted 24 weeks of concurrent training. The exercise intervention did not reduce total cervical adipose tissue, but neck circumference in the vigorous-intensity exercise group decreased by 0.8 cm and 1.0 cm relative to the moderate-intensity group and the control group, respectively (P ≤ 0.05); changes in neck circumference were weakly correlated with body weight, fat mass, and cardiometabolic risk indicators (Arias-Tellez et al., 2023). This finding has two implications. First, exercise training can reduce neck circumference without reducing total cervical fat, suggesting that changes in neck circumference may partly reflect changes in posture, muscle tone, or other soft-tissue components. Second, the study population was young healthy adults rather than patients with buffalo hump or populations at high risk for stroke, and the clinical significance of the observed reduction in neck circumference cannot be directly extrapolated. It provides preliminary evidence relevant to Link 3 but does not address Link 4.

Taken together, the four intervention pathways—surgery, diet, physical therapy, and exercise—all have preliminary evidence at the technical level of reducing the buffalo hump or cervical fat or lowering neck circumference. However, the strength of evidence, target tissue, and population generalizability differ significantly across pathways. The surgical meta-analysis covers the buffalo hump population but does not report neck circumference outcomes (Escandón et al., 2026). The CENTRAL trial has the highest level of evidence, but its target tissue is not the buffalo hump (Tsaban et al., 2021). Physical therapy research evidence is insufficient, with no controlled studies using neck circumference as an outcome included. The ACTIBATE trial reported between-group differences in neck circumference, but the population was young healthy adults (Arias-Tellez et al., 2023). None of these studies directly addresses Link 3—whether intervention produces a clinically meaningful decrease in neck circumference in buffalo hump populations—and none addresses Link 4—whether such a decrease reduces stroke events.

Translation into stroke prevention therefore still faces two gaps that have not yet been filled by clinical data. The first gap is whether intervention causes a clinically meaningful decrease in neck circumference (Link 3). Existing surgical literature focuses on local cosmetic improvement and surgical complications and has not systematically reported neck circumference as an outcome variable (Escandón et al., 2026; Wang et al., 2022). The CENTRAL trial reported changes in cervical fat area but did not report quantitative changes in neck circumference. Physical therapy studies did not provide controlled data with neck circumference as an outcome. The ACTIBATE trial reported between-group differences in neck circumference, but the population was not patients with buffalo hump. The second gap is whether a decrease in neck circumference translates into a reduction in stroke events (Link 4). Even if an intervention can lower neck circumference, long-term follow-up data are needed to demonstrate that this reduction is accompanied by a decreased risk of ischemic stroke incidence or mortality. At present, no direct studies exist for either of these two gaps. For the chain from intervention to reduction in neck circumference to stroke prevention, indirect evidence exists at both the technical starting point and the epidemiologic endpoint. However, the break at the intermediate links means that it currently remains a clinical proposition awaiting testing.

In addition, if forward head posture and posterior cervical mechanical loading independently contribute to part of the risk, then the intervention logic of fat debulking alone also rests on an untested exclusivity premise (Behzadi et al., 2024).

5. Mechanistic Pathways and Preventive Implications

The association between neck circumference and stroke risk is currently supported in the existing literature mainly by three mechanistic explanations that have relatively direct evidence; in addition, the mechanical loading in the posterior cervical region where the buffalo hump is located constitutes a candidate pathway independent of fat accumulation and needs to be listed separately as a competing or supplementary hypothesis. These mechanistic data should be regarded as corroborating or hypothesis-generating evidence; they do not establish that reducing neck circumference reduces stroke events. To distinguish the main chain from supporting evidence, Pathway 1 is presented in two layers: 1A addresses the metabolic activity of upper-body fat relevant to the neck circumference–atherosclerosis link, and 1B addresses pericarotid fat density and perivascular adipose tissue imaging phenotypes as mechanistic corroboration whose quantitative relationship with neck circumference has not been established.

Pathway 1A: Endocrine and Metabolic Activity of Upper-Body Subcutaneous Fat and the Atherosclerosis Pathway. The upper-body subcutaneous fat reflected by neck circumference is generally considered to have metabolic significance independent of overall obesity (Torriani et al., 2014). Free fatty acid release, secretion of proinflammatory adipokines, and the resulting insulin resistance and vascular endothelial dysfunction constitute a classic metabolic pathway from abnormal fat distribution to atherosclerosis. In the Framingham study, neck circumference remained continuously and independently associated with multiple metabolic risk factors after adjustment for visceral fat and BMI, supporting the existence of this pathway (Preis et al., 2010). Systematic reviews and meta-analyses further support, at the cross-study level, the associations of neck circumference with cardiometabolic risk factors and metabolic syndrome (Ataie-Jafari et al., 2018; Namazi et al., 2018). The fat compartmentalization study by Torriani et al. (2014) suggested that the dorsocervical fat compartment where the buffalo hump is located is itself independently associated with metabolic syndrome; the MRI anatomical data of Wang et al. (2022) further supplemented, at the clinical morphologic level, the observation that the dorsocervical fat pad is only weakly correlated with obesity. The ELSA-Brasil study provided position-specific evidence at the vascular imaging level for this pathway: after full adjustment for BMI and waist circumference, neck circumference was independently associated with carotid intima-media thickness but not with coronary artery calcium score (Baena et al., 2016). This dissociation pattern is consistent with paracrine or mechanical effects of local cervical fat on the adjacent carotid artery. Human tissue studies further provided a molecular anchor for this explanation: in patients with carotid stenosis, gene expression of visfatin, monocyte chemoattractant protein-1, and CD68 in pericarotid adipose tissue was significantly higher than that in subcutaneous fat, suggesting that fat depots adjacent to vessels may directly participate in atherogenesis through local paracrine inflammation (Pandzic Jaksic et al., 2022).

Pathway 1B: Pericarotid Fat Density and Perivascular Adipose Tissue Imaging Phenotypes as Mechanistic Corroboration. Between the above molecular evidence and clinical events, the study by Chen et al. (2025) provided a measurable imaging intermediate phenotype. This retrospective study, based on computed tomography angiography (CTA) and magnetic resonance high-resolution vessel wall imaging in patients with carotid plaque, measured pericarotid fat density (PFD), expressed on computed tomography as perivascular adipose tissue (PVAT) attenuation. It found that the area under the receiver operating characteristic curve (AUC) of symptomatic-side PFD for predicting acute ischemic events was 0.762 (95% CI, 0.653–0.870), significantly better than that of contralateral PFD (AUC, 0.672) and that of the inter-side PFD difference (AUC, 0.660). Multivariable regression showed that symptomatic-side PFD was independently associated with intraplaque hemorrhage (β = 0.367; P < 0.001), lipid-rich necrotic core (β = 0.190; P = 0.046), and plaque thickness (β = 0.225; P = 0.027), with the strongest association observed for intraplaque hemorrhage. PFD reflects density changes in the perivascular fat immediately adjacent to the carotid artery. In states of vascular inflammation, the lipid content of perivascular fat decreases and the computed tomography attenuation value increases. As a quantitative indicator of this process, PFD directly corresponds anatomically to the hypothesis in Pathway 1A regarding the local effect of dorsocervical fat on the adjacent carotid artery. The ipsilateral dominance pattern, in which symptomatic-side PFD is superior to contralateral PFD, supports the spatial logic of local fat–local vessel–local event more strongly than a simple cross-sectional association. Mao et al. (2025) supplemented, at the same imaging level, the contemporaneous associations of PFD with circulating inflammation and plaque vulnerability. Among 498 patients who underwent cervical CTA, PFD was positively correlated with C-reactive protein, the neutrophil-to-lymphocyte ratio, the platelet-to-lymphocyte ratio, and the systemic immune-inflammation index, and negatively correlated with the lymphocyte-to-monocyte ratio. After adjustment for systemic inflammatory markers, higher PFD remained independently associated with extracranial internal carotid artery stenosis, with an OR of 1.179 (95% CI, 1.003–1.387), and vulnerable carotid plaque, with an OR of 1.046 (95% CI, 1.021–1.072). This result advances the local proinflammatory phenotype of perivascular fat from histologic gene expression evidence to a level that can be routinely measured in clinical imaging, filling in the contemporaneous association link of PFD–circulating inflammation–plaque vulnerability. However, it was a single-center retrospective design and can only demonstrate co-occurrence rather than infer causality.

At the level of event association, a systematic review and meta-analysis of 8 studies including 2,472 participants showed that carotid perivascular fat attenuation was significantly higher in patients with symptomatic ischemic stroke or TIA than in asymptomatic individuals, with a standardized mean difference (SMD) of 1.10 (95% CI, 0.67–1.53), whereas plaque thickness did not differ significantly between the two groups. This suggests that the local vascular inflammation reflected by this imaging marker is associated with ischemic events at the cross-study level (Khanna et al., 2025).

Beyond the above density and systematic review evidence, longitudinal studies with stronger temporality have provided another layer of support. The longitudinal predictive value of PVAT attenuation for ipsilateral stroke or TIA has been reported in patients with symptomatic carotid stenosis, with an adjusted HR of 3.21 per 10 Hounsfield units (HU) over 5.1 years of follow-up (Bierens et al., 2026). Xu et al. (2025) reported the independent predictive value of PFD for recurrent ischemic stroke or TIA in a more general CTA population. Among 739 patients who underwent CTA with a median follow-up of 3.31 years, 166 reached the recurrence endpoint, and the adjusted HR for PFD was 1.29 (95% CI, 1.02–1.63). Adding PFD to the baseline model increased the C-statistic from 0.61 to 0.84. This study extended the longitudinal predictive evidence for PFD from a highly selected population with symptomatic carotid stenosis to a larger population whose inclusion criteria were closer to real-world CTA examination scenarios, and its temporality is superior to that of existing cross-sectional evidence. However, it remains a single-center retrospective design, and the population was limited to those who underwent CTA; the findings therefore cannot be extrapolated to asymptomatic populations with buffalo hump or the general population with high neck circumference; moreover, the study did not involve neck circumference, and the quantitative relationship between PFD/PVAT and neck circumference has still not been established. In addition, an ipsilateral dominance of PFD relative to the infarct has also been observed in embolic stroke of undetermined source, and there was no significant difference in ipsilateral PFD between embolic stroke of undetermined source and large-artery atherosclerotic stroke (Hu X et al., 2022). This suggests that the local perivascular inflammation marked by this imaging phenotype may not depend on classic large-artery atherosclerotic stenosis; however, the study was a retrospective design, so the causal direction between PFD and embolic stroke of undetermined source cannot be inferred on that basis, and the quantitative relationship between PFD and neck circumference has still not been established.

On this basis, two recent multicenter radiomics studies further advanced PVAT from a mean density measure to a high-dimensional imaging phenotype. Zhao et al. (2025) included 645 patients who underwent carotid CTA at three centers (232 symptomatic and 413 asymptomatic), extracted PVAT radiomics features, and compared seven classifiers. XGBoost performed best in the validation set (mean AUC, 0.797). A combined model integrating the radiomics score (Rad-score) with white blood cell count and hyperlipidemia achieved an AUC of 0.836 in two independent external validation sets and provided calibration curves and decision curve analysis. Sun et al. (2026) focused on 582 patients with asymptomatic carotid stenosis. They constructed a training set, an internal validation set, and two independent external validation sets from three centers; selected 9 PVAT radiomics features; and used a combined model incorporating only three variables—hypertension, soft plaque, and the Rad-score. This model achieved AUCs of 0.911, 0.868, 0.882, and 0.871 across the four datasets, with short-term stroke events during follow-up as the endpoint. The value of these two studies lies in using multicenter external validation and machine learning models to expand PVAT from a single density indicator into a high-dimensional imaging phenotype that can be embedded in clinical prediction frameworks, thereby partially addressing the technical need for integrating incremental predictive value into a unified assessment framework. However, their limitations are equally clear: both were retrospective designs; the endpoint in Zhao et al. was symptomatic plaque rather than a hard stroke endpoint; the population in Sun et al. was limited to asymptomatic carotid stenosis with a follow-up window of only 2 years; neither involved neck circumference; and the quantitative relationship between PVAT radiomics and neck circumference has still not been established. Therefore, these studies strengthen the methodological basis of the imaging intermediate phenotype in Pathway 1B rather than filling the epidemiologic evidence gaps linking buffalo hump to neck circumference or neck circumference to stroke.

In contrast to the above approach of building separate prediction models using PVAT features, Qian et al. (2025) provided another integration logic: embedding PFD into an existing carotid plaque risk stratification tool (plaque-RADS) rather than constructing a new model from scratch. This retrospective study included 272 patients with carotid plaque, with a follow-up of 17.24 months, and 64 patients (23.3%) experienced stroke recurrence. The composite risk index constructed by the authors is a binary variable based on plaque-RADS and PFD: low risk was defined as plaque-RADS < 3 and PFD ≤ −74 HU, and high risk as plaque-RADS ≥ 3 or PFD > −74 HU. The composite risk index itself was an independent predictor of stroke recurrence, with an OR of 8.25 (95% CI, 2.23–30.44); a combined model incorporating degree of stenosis, plaque-RADS, ipsilateral PFD, the bilateral PFD difference, and the composite risk index achieved an AUC of 0.892, and the authors noted that the risk-stratification gain was particularly pronounced in patients with mild to moderate stenosis. The significance of this design for the argument lies in what it answers: not whether PFD has predictive value—a point already established by Chen et al. (2025), Khanna et al. (2025), and Xu et al. (2025)—but whether PFD can provide incremental information on top of existing clinical stratification tools. This is logically analogous to the neck circumference incremental indicators discussed in Section 6 (e.g., the 12.1% improvement in correct classification in Li G et al. (2024) and the net reclassification improvement (NRI) of 0.115 in Li G et al. (2022)), except that the exposure variable is changed from neck circumference to PFD and the existing framework is changed from a general risk model to plaque-RADS. Therefore, the study by Qian et al. (2025) responds more directly to the word integration than do Zhao et al. (2025) and Sun et al. (2026): the latter two build new models using PVAT, whereas the former adds to an existing model. However, its limitations also need to be clearly stated. First, the study was a retrospective design, and the cutoffs and logical combination rules of the composite risk index were determined within the derivation cohort, without independent external validation. Second, the AUC of 0.892 is the absolute discriminative performance of the combined model rather than the incremental AUC of PFD relative to plaque-RADS alone, and the two types of metrics are not directly comparable. Third, the endpoint was stroke recurrence rather than first-ever stroke, and the population was limited to patients with existing carotid plaque. Fourth, the quantitative relationship between PFD and neck circumference was not addressed at all. Therefore, Qian et al. (2025) provide an operationalized example of how PFD can be embedded into the clinical decision-making process rather than yet another piece of association evidence for the predictive performance of PFD.

The association between neck circumference and atherosclerotic burden is not limited to the carotid artery. Although the following study used coronary endpoints, it provides a parallel reference for neck circumference as a marker of systemic atherosclerosis. The cross-sectional study by Yang H et al. (2025) included 1,652 patients who underwent coronary angiography for chest pain and found that, compared with the lowest quartile, the highest quartile of neck circumference was associated with an OR for coronary heart disease of 1.80 (95% CI, 1.22–2.66) and was independently associated with the severity of coronary lesions, with an OR of 1.41 (95% CI, 1.02–1.95). Combined with the dissociation pattern observed in the ELSA-Brasil study—in which neck circumference was associated with carotid intima-media thickness but not with coronary artery calcium score (Baena et al., 2016)—these two studies together suggest that the pathologic process marked by neck circumference may be more consistent with a noncalcified, inflammation-driven atherosclerotic pathway than with a purely calcified lesion. The endpoint of that study was coronary heart disease rather than stroke, and it serves only as a parallel reference in mechanistic inference.

Pathway 2: The Mediating Role of Obstructive Sleep Apnea. The association between neck circumference and OSA has been repeatedly observed, and OSA is a known risk factor for stroke. Li G et al. (2022), in their cross-sectional study, invoked this pathway to explain sex differences, noting that the association between cervical fat and OSA severity was more pronounced in women. The systematic review by Yang H et al. (2024) provided descriptive support for this pathway: among studies of OSA prediction models in patients with stroke, neck circumference, together with BMI, sex, snoring, and blood pressure, was one of the most frequently selected predictors. However, the topic of that review was variable selection for prediction models, and it cannot serve directly as proof of a mediating mechanism. The reliability study by Gjevre et al. (2013) in women with suspected OSA further showed that the interrater ICC for neck circumference was 0.78, and the AUC for predicting an apnea-hypopnea index ≥5 was 0.785, comparable to that of BMI, suggesting that both the measurement operability and the predictive information of neck circumference are clinically acceptable in this high-risk population. The retrospective cohort study by Yao et al. (2025), based on a Chinese population with hypertension and comorbid OSA, provided longitudinal evidence: patients with neck circumference >40 cm had a significantly increased risk of cardiac events, with an HR of 1.39 (95% CI, 1.01–1.93). It should be made clear that the endpoint of that study was cardiac events rather than a composite cerebrovascular endpoint including stroke. Therefore, its position in this review is as corroborating evidence within the OSA pathway that neck circumference still carries prognostic information in high-risk populations, rather than as part of the core neck circumference–stroke evidence in Section 3. The preventive implication of this pathway is that if an intervention can reduce neck circumference and alleviate OSA severity, it may reduce stroke risk by improving nocturnal hypoxemia and sympathetic activation; however, this mediating chain has not yet been directly tested in intervention studies.

Pathway 3: Atrial Fibrillation–Related Cardioembolic Pathway. The Framingham Heart Study followed 4,093 participants without atrial fibrillation at baseline for 11.2 years and found that larger neck circumference was independently associated with incident atrial fibrillation; this association remained significant after adjustment for BMI, with a subdistribution HR (sHR) of 1.51 (95% CI, 1.21–1.89); after adjustment for waist circumference, the sHR was 1.47 (95% CI, 1.18–1.83); and after adjustment for height and body weight, the sHR was 1.37 (95% CI, 1.09–1.72) (Kornej et al., 2022). This suggests that the risk information carried by neck circumference may partly reflect the substrate of atrial structural or electrophysiologic remodeling. The endpoint of that study was atrial fibrillation rather than stroke, so it cannot answer the causal question of whether lowering neck circumference reduces atrial fibrillation or stroke; however, in terms of the mechanistic landscape, it expands the candidate pathways linking neck circumference to stroke from two—atherosclerosis and OSA—to three. No intervention study has yet tested this mediating chain.

Pathway 4: Forward Head Posture and Posterior Cervical Mechanical Loading—A Candidate Pathway Independent of Fat Accumulation. The above three pathways all center on adipose tissue and its metabolic and endocrine activity. However, in addition to fat accumulation, the dorsocervical region where the buffalo hump is located also involves mechanical dimensions related to posterior cervical soft-tissue tension and head-neck posture. In clinical observation, buffalo hump frequently co-occurs with forward head posture, and the two often appear together in the postural pattern of upper crossed syndrome; however, the existing literature still lacks systematic measurements of the prevalence of forward head posture specifically in populations with buffalo hump, as well as postural data stratified by buffalo hump subtype. Therefore, this co-occurrence should be described as a common clinical combination rather than a quantitative rule, and Pathway 4 should be regarded as an untested candidate pathway rather than an established mechanism.

At the level of clinical endpoints, the matched case-control study by Behzadi et al. (2024) included 206 patients with cervicalgia (103 with posterior circulation ischemic stroke versus 103 controls), matched for known stroke risk factors such as hypertension, hyperlipidemia, diabetes, smoking, hypercoagulable state, and atrial fibrillation. The study found that the C0–C2 angle was significantly larger in the stroke group than in the control group (30.7° versus 23.2°; P = 0.002), and the sagittal vertical axis offset was also significantly greater (35.4 mm versus 25.3 mm; P < 0.001). The authors interpreted this as suggesting that chronic changes in vertebrobasilar circulatory hemodynamics under forward head posture may contribute to posterior circulation ischemia.

This study provides the human clinical association evidence currently available in the published literature for forward head posture–posterior circulation ischemic stroke. However, its evidentiary boundaries must be stated at the same time. First, a case-control design cannot establish temporality; the temporal relationship between forward head posture and stroke cannot be determined on this basis, and reverse causation—post-stroke changes in posture—cannot be excluded. Second, the study population was limited to patients with cervicalgia and cannot be directly extrapolated to asymptomatic populations with buffalo hump. Third, the C0–C2 angle and sagittal vertical axis offset mainly reflect the degree of forward inclination at the craniocervical junction and upper cervical spine, whereas the buffalo hump is located at the cervicothoracic junction (C7–T1 region); there is therefore an anatomical gap in localization between the two. Fourth, the study did not measure the buffalo hump, neck circumference, or cervical fat mass and therefore cannot demonstrate that the association with forward head posture is independent of fat accumulation.

Its function in the argument of this article is not to add another mechanistic pathway with established evidentiary depth but rather to highlight an untested exclusivity premise. If forward head posture or posterior cervical mechanical loading does independently affect posterior circulation risk beyond fat mass, then reducing buffalo hump fat solely through surgery, diet, or exercise would not necessarily eliminate all mechanical pathway risk. None of the existing intervention studies (Escandón et al., 2026; Tsaban et al., 2021; Arias-Tellez et al., 2023; Wang et al., 2022) concurrently measured forward head posture or distinguished the respective contributions of fat reduction and postural improvement. Pathway 4 should therefore be positioned as an independent line of evidence that compels the existing fat-centered framework to undergo testing rather than as a closed causal link.

6. From Risk Marker to Prevention Strategy: Structural Gaps and a Research Agenda

The descriptive synthesis above indicates that the strength of evidence differs markedly across the four links. First, the starting point of the chain is inferential. The evidence linking neck circumference to stroke comes from studies in general populations, and neck circumference was not measured as a surrogate marker for buffalo hump. The quantitative relationship between buffalo hump and neck circumference—including whether the contribution of each buffalo hump subtype to neck circumference differs—has not been specifically studied (Li FW et al., 2026); Lee et al. (2025) also acknowledged in their non-HIV buffalo hump cohort that neck circumference data were not systematically collected. Translating the epidemiologic evidence for neck circumference to populations with buffalo hump requires a prerequisite validation that has not yet been completed. This is the core gap in Link 1 of the four-link framework—from buffalo hump to neck circumference. Although buffalo hump and forward head posture frequently co-occur in clinical observation, this co-occurrence likewise lacks systematic quantitative measurement in populations with buffalo hump and therefore cannot serve as an established intermediate link connecting buffalo hump to stroke.

Second, Link 3 remains incompletely tested and Link 4 remains untested: no study has reported a clinically meaningful postintervention reduction in neck circumference in buffalo hump populations together with stroke events. The surgical meta-analysis confirmed the technical feasibility of debulking (Escandón et al., 2026). The case series by Wang et al. (2022) reported local cosmetic improvement. The CENTRAL trial confirmed that dietary intervention can reduce cervical fat (Tsaban et al., 2021), and the ACTIBATE trial suggested that exercise can lower neck circumference in young adults (Arias-Tellez et al., 2023). However, none of these studies systematically reported the quantitative change in neck circumference in patients with buffalo hump after intervention. Studies by Li G et al. (2024), Cao et al. (2022), BCDS-20 (Yang GR et al., 2019), and Han and Kim (2022) confirmed the association of neck circumference with cerebrovascular outcomes or hard endpoint events including stroke, but did not involve intervention. The two lines of evidence are disconnected at Link 3—whether intervention causes a clinically meaningful decrease in neck circumference; even if this gap were filled, Link 4—whether a decrease in neck circumference is accompanied by a reduction in stroke events—would still be required to complete the full argument from intervention to prevention. At present, no direct data exist for either of these two gaps. In addition, if forward head posture or posterior cervical mechanical loading independently contributes to part of the risk, then the intervention logic of fat debulking alone also rests on an untested exclusivity premise (Behzadi et al., 2024).

Third, the sources of population heterogeneity and effect modification by sex have not yet been explained. The direction of effect modification by sex is also not uniform. The Jackson Heart Study and the Framingham Offspring Cohort reported negative results, whereas East Asian populations showed relatively consistent positive findings (Pumill et al., 2019; Rosenquist et al., 2013). Li G et al. (2022) and Han and Kim (2022) further observed more robust associations in women, whereas Hu T et al. (2022), using the same cutoffs in a neighboring population, observed stronger associations in men. These contrasts suggest that effect modification may arise from racial and sex differences in fat distribution, differences in the relative mediating contribution of OSA, differences in the relative contribution of the atrial fibrillation pathway, or other unmeasured pathways. Whether these differences partly reflect effect modification by climatic background (e.g., seasonal effects of cold exposure on blood pressure and vascular tone) has not yet been directly examined in neck circumference populations; the contrast in geographic and sex patterns between Hu T et al. (2022), on the one hand, and Li G et al. (2022) and Han and Kim (2022), on the other, is only an ecological clue and does not constitute evidence. The predictive value of neck circumference observed by Yao et al. (2025) in patients with hypertension and comorbid OSA, together with the association between neck circumference and atrial fibrillation observed by Kornej et al. (2022) in the Framingham cohort, suggests that comorbidity status and arrhythmic endpoints may be another source of effect modification. The opposite predictive directions of neck circumference and BMI observed by Li Y et al. (2021) in patients with intracerebral hemorrhage, together with the proinflammatory phenotype of perivascular fat observed by Pandzic Jaksic et al. (2022) in patients with carotid stenosis, further suggest that the pathophysiological meaning of neck circumference may differ across vascular beds and stroke subtypes. With respect to measurement methodology, the reliability data of Sato et al. (2024), Hagren et al. (2023), and Gjevre et al. (2013), together with the systematic review by Espinoza López et al. (2021) on inconsistent measurement sites and cutoff definitions, suggest that measurement variation in cross-cohort comparisons is at least a competing source of explanation beyond biological differences. In the absence of mechanistic studies, this heterogeneity is better regarded as a variable that future prevention research designs should consider through stratification rather than being assigned a definitive pathophysiological meaning. Forward head posture, as another potential effect modifier or independent pathway, should also be considered in parallel with stratification by anterior versus posterior circulation stroke subtype in future research (Behzadi et al., 2024).

Fourth, the incremental predictive value of neck circumference has been quantified in several studies, but these data have not yet been integrated into a unified assessment framework. The prospective cohort of Li G et al. (2024) showed that adding neck circumference to an existing model improved the correct classification rate for overall cardiovascular mortality by 12.1%. The cross-sectional study of Li G et al. (2022) showed that adding neck circumference in women improved the correct classification rate for ischemic stroke by 11.0% (NRI 0.115). The study of Yang H et al. (2025) showed that neck circumference combined with the atherogenic index of plasma achieved an AUC of 0.729 for predicting coronary heart disease. BCDS-20 reported an adjusted HR of 2.305 in the type 2 diabetes population, providing a reference at the level of hard endpoint events for incremental value, but that study did not report incremental metrics such as NRI or AUC (Yang GR et al., 2019). Recent multicenter PVAT radiomics studies by Zhao et al. (2025) and Sun et al. (2026) demonstrated another integration pathway. These studies constructed externally validated models using the Rad-score combined with a small number of clinical variables, achieving relatively high AUCs in symptomatic plaque identification or short-term stroke prediction in asymptomatic carotid stenosis. Qian et al. (2025) provided a third integration logic. Rather than building a separate model, they embedded PFD into the existing plaque-RADS scoring system and demonstrated incremental value in predicting stroke recurrence using a binary composite risk index. It should be specifically noted that the AUC of 0.892 reported by Qian et al. (2025) represents the absolute discriminative performance of the combined model rather than the incremental AUC of PFD relative to plaque-RADS alone, and the two types of metrics are not directly comparable. However, their approach of adding to an existing stratification tool is closer to the actual form of clinical decision-making than simply constructing a new radiomics model. These three types of incremental value data differ in nature, population, and endpoint: incremental metrics from adding neck circumference to a general risk model, the absolute discriminative performance of PVAT radiomics models, and the composite risk index from embedding PFD into plaque-RADS. At present, a unified assessment in buffalo hump or high neck circumference populations with stroke as the endpoint is still lacking; cross-study dose-response evidence also comes mainly from within single cohorts, and a meta-analysis dedicated to the neck circumference–stroke association has not yet been conducted.

Fifth, translating risk markers into prevention strategies requires interventional evidence. Observational studies can establish an association between neck circumference and stroke risk but cannot answer whether lowering neck circumference can prevent stroke. To demonstrate the stroke-preventive value of intervention, prospective intervention cohorts are needed. Such cohorts would implement surgical debulking, dietary intervention, physical therapy, or exercise training in populations with buffalo hump or high neck circumference; systematically measure preoperative and postoperative neck circumference (and report effect sizes and confidence intervals); and longitudinally track ischemic stroke events. The feasibility of this research agenda depends on whether several variables can be collected concurrently within a single study: buffalo hump subtype; neck circumference (using a standardized measurement protocol with reliability control); carotid imaging parameters (including intermediate phenotypes such as PFD, PVAT radiomics features, and plaque-RADS combined with PFD; Zhao et al., 2025; Sun et al., 2026; Qian et al., 2025); OSA severity; incident atrial fibrillation; forward head posture parameters; and stroke endpoints. Forward head posture is included not because it is an established independent pathway but to test whether the exclusivity premise that fat debulking alone is sufficient to reduce risk holds (Behzadi et al., 2024).

7. Conclusions

The existing literature suggests that an increase in neck circumference is associated with the risk of ischemic stroke and its subclinical precursor (carotid atherosclerosis) independently of overall obesity. Between mortality endpoints and subclinical endpoints, BCDS-20 provided prospective evidence on the incidence of cardiovascular events including stroke in 3,009 patients with type 2 diabetes (adjusted HR of approximately 2.3), thereby providing direct support for the longitudinal chain from neck circumference to a composite hard endpoint that included stroke in a high-risk diabetic population (Yang GR et al., 2019); this estimate was not stroke-specific. The association was relatively robust in women in some East Asian cohorts (Li G et al., 2022; Han and Kim, 2022), but Hu T et al. (2022), in a 7.6-year prospective cohort from the same geographic region and using the same neck circumference cutoffs, reported stronger predictive performance in men, suggesting that the direction of effect modification by sex is not consistent across cohorts; negative results from the Jackson Heart Study in African Americans and the Framingham Offspring Cohort in White individuals indicate that this association is not universally present (Pumill et al., 2019; Rosenquist et al., 2013). Morphologic classification studies have shown that buffalo hump tissue composition is more heterogeneous than simple fat accumulation (Li FW et al., 2026). Clinical data from non-HIV populations further indicate that buffalo hump frequently co-occurs with obesity, glucose and lipid metabolism disorders, and the severity of liver histologic lesions (Lee et al., 2025; Cheung et al., 2007). Cervical fat compartmentalization studies have provided imaging clues for the independent metabolic effect of local dorsocervical fat (Torriani et al., 2014), and MRI anatomical data have supplemented its localization and dimensions while reporting only a weak correlation with obesity (Wang et al., 2022). Buffalo hump frequently co-occurs with forward head posture in clinical observation, but this co-occurrence remains a postural-pattern observation rather than a systematically quantified relationship in buffalo hump populations.

At the mechanistic level, the position-specific findings of the ELSA-Brasil study support a selective effect of local cervical fat on the adjacent carotid artery (Baena et al., 2016), and the proinflammatory gene expression of perivascular fat in patients with carotid stenosis provides a molecular anchor for this position specificity (Pandzic Jaksic et al., 2022). Several lines of evidence have advanced this pathway from molecular clues to measurable imaging intermediate phenotypes and have partially filled the temporal dimension, including the independent association between symptomatic-side PFD and acute ischemic stroke (AUC, 0.762) and the strong association between PFD and intraplaque hemorrhage (Chen et al., 2025); the contemporaneous associations of PFD with circulating inflammatory markers and plaque vulnerability (Mao et al., 2025); the cross-study pattern in which carotid perivascular fat attenuation was elevated in symptomatic patients (SMD, 1.10) while plaque thickness did not differ significantly (Khanna et al., 2025); the ipsilateral dominance of PFD relative to the infarct in embolic stroke of undetermined source (Hu X et al., 2022); the longitudinal predictive value of PVAT attenuation for ipsilateral stroke or TIA (Bierens et al., 2026); and the independent predictive value of PFD for recurrent ischemic stroke or TIA in a more general CTA population (Xu et al., 2025). The multicenter radiomics studies by Zhao et al. (2025) and Sun et al. (2026) further expanded PVAT from density measurement to externally validated high-dimensional imaging phenotypes, whereas Qian et al. (2025) embedded PFD into the existing plaque-RADS scoring system and demonstrated incremental value in predicting stroke recurrence through a composite risk index. Together, these three studies provide a methodological spectrum for an integrative framework, ranging from building separate models to embedding into existing tools; however, their retrospective designs, endpoint definitions, and lack of a quantitative relationship with neck circumference still limit their translation into prevention strategies. The quantitative relationship between PFD/PVAT and neck circumference has not yet been established, and the predictive performance of imaging markers cannot be equated with the effect of neck circumference.

Regarding intervention, surgical debulking is technically feasible (Escandón et al., 2026); the case series by Wang et al. (2022) reported local cosmetic improvement but did not report neck circumference; the dietary intervention randomized controlled trial confirmed that cervical fat can be reduced with accompanying metabolic improvement (Tsaban et al., 2021); and the exercise intervention randomized controlled trial showed that the reduction in neck circumference may not depend on a reduction in total cervical fat mass (Arias-Tellez et al., 2023). Mechanistically, the literature base for OSA as a candidate mediating pathway has expanded from single-population studies to the level of systematic reviews (Yang H et al., 2024); Gjevre et al. (2013) reported an interrater ICC of 0.78 for neck circumference and an AUC of 0.785 for predicting apnea-hypopnea index ≥5 in women with suspected OSA; and Yao et al. (2025) obtained longitudinal evidence for a cardiac event endpoint in a population with hypertension and comorbid OSA. The Framingham Heart Study suggested that neck circumference is independently associated with the risk of incident atrial fibrillation, providing a clue for the cardioembolic pathway (Kornej et al., 2022). Prognostic studies after acute ischemic stroke suggest that neck circumference may have clinical significance in post-stroke mortality risk assessment (Medeiros et al., 2011); a study of functional outcomes after spontaneous intracerebral hemorrhage further suggested that the pathophysiological information reflected by neck circumference and BMI differs fundamentally (Li Y et al., 2021); and the association between neck circumference and coronary heart disease supports its interpretation as a surface marker of systemic atherosclerosis from the perspective of another vascular bed (Yang H et al., 2025). Behzadi et al. (2024), in a case-control study of patients with cervicalgia, suggested that forward head posture is associated with posterior circulation ischemic stroke, providing indirect human clinical evidence that mechanical loading in the posterior cervical region where the buffalo hump is located may constitute a candidate pathway independent of fat accumulation; however, this evidence cannot establish temporality, cannot be extrapolated to populations with buffalo hump, has not demonstrated independence from fat mass, and cannot exclude reverse causation. Multiple studies have quantified the incremental predictive value of neck circumference, but these data are scattered across different populations and endpoints and have not yet been integrated into a unified risk-stratification framework with stroke as the endpoint.

Efforts to integrate the above evidence into a complete strategy from buffalo hump to neck circumference and from intervention to stroke prevention still face two central translational gaps. First, Link 1 remains inferential: the quantitative relationship between buffalo hump and neck circumference lacks systematic observation within the same population, and an increase in buffalo hump accompanied by an increase in neck circumference is currently an anatomical inference rather than a conclusion from direct measurement (Li FW et al., 2026; Lee et al., 2025); the frequent co-occurrence of buffalo hump with forward head posture likewise lacks systematic verification in populations with buffalo hump. Second, Link 3 remains incompletely tested and Link 4 remains untested: no interventional or longitudinal study has reported a clinically meaningful postintervention reduction in neck circumference in buffalo hump populations together with stroke events. BCDS-20 (Yang GR et al., 2019) and the studies by Chen et al. (2025), Mao et al. (2025), Khanna et al. (2025), Hu X et al. (2022), Bierens et al. (2026), Xu et al. (2025), Zhao et al. (2025), Sun et al. (2026), and Qian et al. (2025) have strengthened the intermediate links of the argument from multiple directions—incidence of hard endpoint events, imaging intermediate phenotypes, contemporaneous inflammatory associations, longitudinal prediction, radiomics integration, and incremental models embedded in existing stratification tools; Hu T et al. (2022) used the inconsistency in the direction of effect modification by sex to suggest that the boundaries of the existing epidemiologic evidence are more complex than previously assumed; Sato et al. (2024), Hagren et al. (2023), Gjevre et al. (2013), and Espinoza López et al. (2021) suggested, from the perspective of measurement reliability and variability in measurement location, that cross-cohort comparability itself constitutes a methodological limitation; and Behzadi et al. (2024) suggested that a mechanical candidate pathway that has not yet been tested exists outside the fat-centered framework. None of these studies, however, changes one basic judgment: the existing evidence supports treating buffalo hump and neck circumference as surface markers of stroke risk and potential intermediate targets, but it is still insufficient to support the clinical conclusion that treating or reducing buffalo hump, or lowering neck circumference, can prevent stroke.

If future research can concurrently collect neck circumference (using a standardized measurement protocol and reporting reliability), buffalo hump subtype, carotid imaging parameters (including intermediate phenotypes such as PFD/PVAT, radiomics features, and plaque-RADS combined with PFD), OSA severity, incident atrial fibrillation, forward head posture parameters, and stroke endpoints in populations with buffalo hump or high neck circumference, and systematically track postintervention neck circumference trajectories and stroke events, it may be possible to translate the current descriptive associations into testable prevention hypotheses.