General attributes of the study sample

The demographic characteristics and laboratory data of the included individuals with or without migraine are summarized in Table 1. We grouped these participants based on whether they had migraine and then compared their baseline conditions.

Table 1 General characteristics of the study participants (n = 3,354) in NHANES 1999–2004

After a rigorous screening process, 3354 participants who were older than 20 years and younger than 60 years were enrolled in this study, of which 839 had severe headache or migraine. The overall crude weighting prevalence of migraine was 24.76% (95% CI: 23.09–26.51), and patients are inclined to be females [65.83% (95% CI: 62.00–69.47) in females vs. 34.17% (95% CI: 30.53–38.01) in males]. Furthermore, those participants with severe headache or migraine tended to lower education levels, more divorced, lower income, fewer alcohol consumption, more sedentary behavior (all P < 0.05), and more probable to have elevated CRP levels (0.47 ± 0.04 vs. 0.37 ± 0.02), WC (96.02 ± 0.69 vs. 94.56 ± 0.37), BMI (28.85 ± 0.28 vs. 27.63 ± 0.14), LAP (51.87 ± 1.81 vs. 46.12 ± 0.92), CMI (0.65 ± 0.01 vs. 0.69 ± 0.02), VAI (2.00 ± 0.06 vs. 1.78 ± 0.03), BRI (5.00 ± 0.09 vs. 4.57 ± 0.05), WTI (6.91 ± 0.02 vs. 6.86 ± 0.01), WHtR (0.57 ± 0.00 vs. 0.55 ± 0.00), WWI (1.21 ± 0.01 vs. 1.19 ± 0.00) (all P < 0.05), with higher occurrence of mental health problems [3.19% (95% CI: 2.50,4.06) vs. 6.32% (95% CI: 4.71,8.42)], cancer [6.40% (95% CI: 4.60,8.83) vs. 3.61% (95% CI: 2.84,4.58)] (all P < 0.05). However, there are no discernible variations in the CI and TyG indices between participants with and without migraine (P = 0.12, 0.13, respectively).

The association of lipid-related obesity metrics with severe headache or migraine

The partial correlation among lipid-related obesity indicators and migraine both in continuous and categorical analyses are illustrated in Table 2. As the continuous analysis demonstrated, positive associations were consistently found between WHtR, BRI, BMI, LAP, WTI, VAI and headache in Models 1–3 (all P < 0.05). A strong association between the prevalence of migraines and WWI was observed in Models 1 [odds ratio (OR) = 2.52, 95% CI: 1.31–4.85, P < 0.01], however, this association was not stable both in model 2 and model 3 (P = 0.26, 0.05, respectively). Furthermore, there were no relationships have been found between TyG, CI, CMI and severe headache or migraine in model1 (all P > 0.05). These results finally demonstrated that TyG, CI, CMI cannot perform well in differentiating or forecasting migraine. Among the aforementioned measures, WHtR exhibited the strongest diagnosis ability for migraine (OR = 5.77, 95% CI: 1.93–17.26, P < 0.01, in model 3).

Table 2 The associations of the quartiles of lipid-related obesity indicators with severe headache or migraine

As illustrated by the categorical analysis, in comparison with the first quartile (Q1), which was used as a control group, the significant associations between WHtR, BRI and migraine persisted in the other three quartiles (Q2–Q4). Moreover, with the escalation of WHtR and BRI, the OR values for severe headache or migraine likewise increased (all P for trend < 0.01). Parallelly, BMI, LAP, and WTI were intimately tied to escalated headache risk in the fourth quartile (Q4). More precisely, Q4 of WHtR (OR = 1.68, 95% CI: 1.26–2.25, in model 3), BRI (OR = 1.68, 95% CI: 1.26–2.25, in model 3), BMI (OR = 1.65, 95% CI: 1.26–2.16, in model 3), LAP (OR = 1.54, 95% CI: 1.16–2.05, in model 3) and VAI (OR = 1.36, 95% CI: 1.02–1.80, in model 3) exhibited a significant positive relationship with severe headache or migraine in comparison to Q1–3. In conclusion, higher levels of those five obesity-related indicators are independent risk factors for severe headache or migraine.

Intriguingly, while the continuous analysis indicated a positive linkage from WTI to severe headache or migraine, the categorical analysis revealed insignificant association in any of the quartiles in model 1, suggesting the instable associations between WTI and migraine. Besides, in spite of the continuous analysis demonstrated no association between CMI and migraine, CMIQ3 (CMI 0.54–0.94) had a noticeable association with migraine in Models 1–3.

In addition, stratified analysis by age revealed that WHtR, BRI, BMI, LAP, WTI, VAI, TyG, and CI could increase the risk of developing migraine for those aged 20 to 40 (unadjusted model OR = 37.56, 95% CI: 9.85–143.16; OR = 1.17, 95% CI: 1.10–1.24; OR = 1.05, 95% CI: 1.03–1.07; OR = 1.01, 95% CI: 1.00–1.01; OR = 1.30, 95% CI: 1.06–1.59; 1.12, 95% CI: 1.03–1.22; OR = 1.25, 95% CI: 1.01–1.54; OR = 1.18, 95% CI: 1.03–1.37, respectively, all P < 0.05). Additionally, positive relationship have been found between WWI and migraine among individuals aged 40 to 60 (unadjusted model OR = 3.00, 95% CI: 1.10–8.15, P = 0.03). However, there was no association between CMI and migraine in analyses stratified by age, consistent with the results of the continuous analyses (Supplemental material Table 3).

Following gender stratification, the OR values of the connections between WHtR, BRI, BMI, LAP, WTI, VAI, TyG and CMI could increase the potential for developing severe headache or migraine in females (OR = 4.58, 95% CI: 1.23–7.01; OR = 1.06, 95% CI: 1.00–1.12; OR = 1.02, 95% CI: 1.01–1.04; OR = 1.01, 95% CI: 1.00–1.01; OR = 1.37, 95% CI: 1.10–1.71; OR = 1.10, 95% CI: 1.01–1.19; OR = 1.33, 95% CI: 1.05–1.68; OR = 1.35, 95% CI: 1.05–1.73, in model 3, respectively, all P < 0.05) (Supplemental material Table 3).

Results of stratified analysis and interaction analysis

Stratified analysis and interaction analysis were used to further investigate the role of covariables and lipid-related obesity measures (WHtR, BMI, LAP, WTI, BRI, VAI, WWI) on migraine (Fig. 3).

Fig. 3

Subgroup and interaction analyses of the association of lipid-related obesity indicators and severe headache or migraine

In the stratified analysis, which was stratified by race, WHtR, BMI, BRI, WTI, LAP were substantially associated with migraine among non-Hispanic White people (OR: 6.73, 95% CI: 1.70–26.69; OR: 1.03, 95% CI: 1.01–1.05; OR: 1.08, 95% CI: 1.02–1.15; OR: 1.32, 95% CI: 1.06–1.64; OR: 1.00, 95% CI: 1.00–1.01, respectively). Statistically significant associations were observed between WHtR, BMI, VAI, BRI, LAP and WTI with severe headache or migraine in individuals who were married, higher educational level, higher income. Additional examination of interactions showed that WHtR, BMI, BRI and WTI significantly influenced the risk of migraines in relation to age, race, sedentary behavior and history of cancer (P for interaction < 0.05). Additionally, significant interactions were identified between VAI, LAP and age, race, history of cancer (P for interaction < 0.05). These results could partly explain why WWI were not any more relevant with migraine in Model 2–3.

Nonlinear associations between lipid-related obesity metrics and the prevalence of migraine

In combination with the positive outcomes of the logistic regression analyses, we employed RCS to visually demonstrate dose–response relationships the associations between WHtR, BMI, LAP, WTI, BRI, VAI, WWI and the prevalence of migraine in model1, and the median value of those seven indicators (WHtR = 0.56, BMI = 27.14, LAP = 36.20, WTI = 6.86, BRI = 4.44, VAI = 1.42, WWI = 1.19) were used as the reference point (Fig. 4).

Fig. 4

Restricted cubic splines of lipid-related obesity indicators and severe headache or migraine. Note: The dashed lines represent the 95% confidence interval; the red lines represent inflection points.

Among those seven indicators, WHtR, BMI and BRI demonstrated dose-response relationships with the prevalence of migraine (P for overall = 0.04, 0.01, 0.04, respectively). However, dose-response relationships between LAP, WTI, VAI, WWI and migraine are not significant (P for overall = 0.13, 0.26, 0.25, 0.10, respectively). Furthermore, the RCS plot demonstrated linear relationships between WHtR, BMI and BRI and the risk of migraine (P for non-linearity = 0.20, 0.06, 0.11, respectively), more precisely, the likelihood of experiencing severe headache or migraine rises as BMI, BRI, and WHtR increase.

ROC curves of lipid-related obesity indicators in relation to severe headache or migraine

The ROC curve indicated that WHtR and BRI had comparable and highest diagnostic efficacy for severe headache or migraine (AUC: 0.55, 95% CI: 0.53–0.58; AUC: 0.55, 95% CI: 0.53–0.58), slightly higher specificity compared to BMI (AUC: 0.54, 95% CI: 0.52–0.57) (Fig. 5).

Fig. 5

Receiver operating characteristic curves of WHtR, BMI, BRI and severe headache or migraine