Ultrasound-Guided Greater Occipital Nerve Block Across Headache Phenotypes: Outcomes and Determinants of Response in a Prospective Cohort

Abstract

Introduction

Greater occipital nerve blocks (GONBs) are used across headache phenotypes, but short-term effectiveness and predictors of response in mixed real-world cohorts remain uncertain.

Methods

We conducted a prospective observational study of adults with chronic headache phenotypes receiving a single ultrasound-guided proximal (C2) GONB. Primary outcomes were T0 → T1 (1-month) changes in pain (NRS), headache impact (HIT-6), sleep (PSQI), and monthly attacks (crises) using paired Wilcoxon tests. We defined HIT-6 responders as ≥ 10-point reduction. Predictors of response were evaluated with multivariable logistic regression; continuous reductions were modeled with multivariable linear regression.

Results

Twenty-nine patients were included (chronic migraine n = 10; cluster n = 5; occipital neuralgia n = 7; cervicogenic n = 7; 69% female; mean age 58.2 ± 13.7 years). At 1 month, all outcomes had improved significantly: HIT-6 median 66 → 52 (Δ − 12; 95% CI − 15 to − 8; p < 0.001), PSQI 12 → 8 (Δ − 3.5; p < 0.001), monthly (crises) 24 → 12 (Δ − 14.0; 95% CI − 20.0 to − 10.0; p < 0.001), and NRS 8 → 4 (Δ − 3; p < 0.001). In the logistic model, higher baseline pain independently predicted HIT-6 response (OR 3.38, 95% CI 1.22–13.47; p = 0.017), whereas symptom duration was not associated. In multivariable linear models, cluster headache was associated with greater improvements (ΔNRS, ΔHIT-6, Δcrises), while a rheumatic comorbidity predicted smaller improvements across outcomes.

Conclusion

Across chronic headache phenotypes, a single ultrasound-guided proximal (C2) GONB was associated with clinically meaningful 1-month improvements in pain, impact, sleep, and attack frequency. Response, however, was not uniform: higher baseline pain intensity and the cluster phenotype identified the patients most likely to benefit, whereas a rheumatic comorbidity marked probable non-responders. These predictors offer a practical basis for patient selection and should be confirmed prospectively.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1007/s40122-026-00881-4.

Key Summary Points

Why carry out this study?

Chronic headache is highly disabling, and many patients remain refractory to drug treatment

Ultrasound-guided greater occipital nerve block (GONB) is a low-cost, minimally invasive option, but its short-term effect and predictors of response across headache types remain unclear

We asked whether a single C2 GONB improves short-term outcomes across mixed headache diagnoses and which baseline features predict response

What was learned from the study?

One month after a single GONB, pain (NRS 8 → 4) and headache impact (HIT-6 66 → 52) improved, with better sleep, fewer monthly attacks, and no adverse events

Response varied: higher baseline pain and cluster headache predicted greater benefit and a rheumatic comorbidity predicted less; these predictors are exploratory and need prospective validation

Introduction

Headache disorders are among the most prevalent and disabling neurological conditions worldwide. The Global Burden of Disease study ranks headache disorders as the second leading cause of years lived with disability, with migraine alone the third leading cause among adults aged 15–49 years [1], and they have substantial impacts on productivity, social functioning, family life, and sleep [2]. Headaches and sleep influence each other in complex ways, with significant individual variation, and 30% of patients report headaches related to sleep problems, such as insufficient or excessive sleep (insomnia and hypersomnia) [3].

Given their complex and frequently chronic or relapsing course, management typically requires multimodal care spanning psychological, pharmacological, and interventional options tailored to patient needs [4].

In recent years, among interventional therapies, ultrasound-guided greater occipital nerve block (GONB) has become established as an effective and safe approach to chronic primary headaches, i.e., chronic migraine (CM) or cluster headache (CH), occipital neuralgia (ON), and some secondary headaches [cervicogenic headache (CeH), post-dural puncture headache] owing to low costs, minimal invasiveness, and few pharmacological interactions with other drugs [5].

However, there is still no unanimity regarding the methods for performing the block (blind technique, ultrasound-guided or guided by computed tomography), the level of execution (higher or lower than the C2 level), which drugs to use [corticosteroids or local anesthetics (LA), or a combination], and the need to perform a unilateral or bilateral block, as well as the real effectiveness of the block in terms of pain and crisis reduction [6].

The greater occipital nerve (GON) is a mixed but almost totally sensitive nerve whose innervation periphery reaches superiorly up to the vertex of the head and laterally up to the temporal region, respectively falling within the competence of the trigeminal branches V1 (ophthalmic) and V3 (mandibular), while inferiorly it covers the superior nuchal and mastoid regions [7]. It receives afferents directly from the splenius capitis and semispinalis muscles and, via the anastomotic plexus of Cruveilhier, from the atlanto-axial joint, spinal dura mater of C1 (through the small contingent of sensitive fibers coming from the dorsal branch of C1), and by the longus capitis and multifidus muscles (through the dorsal branch of C3). The axons of the pseudounipolar sensory neurons located in the C1-C2-C3 spinal ganglia are in communication with the axons of the sensory relay neurons of the trigeminocervical nucleus according to a pattern known as “trigeminocervical functional convergence.” In contrast, no such connections are observed between the trigeminal nerve and the C4 spinal nerve or lower spinal nerves. This convergence pattern forms the basis of the referred pain mechanism between the innervation territory of the first three cervical nerves and the trigeminal periphery, explaining the potential role of a GONB in several types of headache, including migraine, cluster, cervicogenic occipital, and orthostatic [8].

The present observational study aims to quantify short-term (T0 → T1) changes in pain, headache impact, sleep, and attack frequency after C2-targeted ultrasound-guided GONB across diagnoses and to identify independent predictors of response.

Methods

This study was performed in accordance with the ethical principles of the Declaration of Helsinki of 1964 and its later amendments and with the International Association for the Study of Pain (IASP) guidelines for pain research in humans. The study protocol was approved by the local ethics committee (Comitato Etico Locale “Gabriella Serio,” IRCCS Istituto Tumori Giovanni Paolo II, Bari, Italy; protocol/IRB no. C2832/CEL—STUDIO NGO 01). Written informed consent to participate was obtained from all individual participants. No identifiable individual patient data are reported; therefore, consent for publication was not applicable. The study was conducted at the Pain Clinic of Policlinico Hospital in Bari from January 2023 to February 2024.

All adult patients referred to the pain clinic of the Policlinico Hospital of Bari with a confirmed diagnosis of primary or secondary chronic refractory headache as per the ICHD-3 classification [9] (International Classification of Headache Disorders, third edition) were screened. Refractory headache was defined as a chronic headache unresponsive to pharmacological treatment. Refractory migraine was defined by having failed all of the available preventatives and suffering from at least eight debilitating headache days per month for at least six consecutive months. Drug failure may include a lack of efficacy or tolerability [10].

The exclusion criteria were allergy to LA, injection site infections, exposed wounds in the affected region, congenital or acquired coagulopathies, Arnold-Chiari malformation [11], and refusal to undergo the block.

All patients, after informed consent, underwent GON blockade using an ultrasound technique at C2 [12]. The procedure was performed with the patient in the prone position using a high-frequency linear probe (7–12 MHz). The scan was carried out in all patients according to a standardized technique, starting from the external occipital protuberance on the midline and progressively descending until it meets the spinous process of the axis, identifying the second cervical vertebra, which is easily recognizable because it is composed of a left and a right tubercle. By slightly rotating the probe obliquely with the end pointing toward the mastoid process of the occipital bone to visualize the axis plate (hyperechoic), located in front of the inferior oblique muscle of the head (hypoechoic), the GON comes as a hypoechoic punctiform structure situated between the semispinalis muscle and the inferior oblique of the head, which appear as two hypoechoic longitudinal structures. Once the GON was located using ultrasound landmarks with the probe placed obliquely, the 27-G needle was inserted in a lateral-to-medial direction following the in-plane technique until the tip reached the target. The injection of the solution was confirmed by fluid distension around the GON (Fig. 1). The rules of asepsis were respected throughout the procedure. The block was performed on the same side of the referred pain area or bilaterally if the patient reported bilateral pain [13]. All blocks were performed by the same operator.

Fig. 1.

Ultrasound image of the left greater occipital nerve block at the C2 level. GON greater occipital nerve, OCI obliquus capitis inferior muscle, SSc semispinalis capitis muscle, C2 lamina lamina of the second cervical vertebra; asterisk, left tubercle of the C2 vertebra; star, interfascial fat plane between the semispinalis capitis and obliquus capitis inferior muscles (injection target)

A solution containing 0.5 ml of LA (levobupivacaine 0.5%) associated with 20 mg of methylprednisolone depot was used, for a total of 1 ml. In case of bilateral block, a total solution of 2 ml was used, 1 ml for each side, and two blocks were performed during the same session. Patients were clinically monitored for 15 min to identify any early side effects. Follow-up control was then set at 4 weeks. During the 1-month follow-up, patients continued their pre-existing headache pharmacotherapy without modification; no additional preventive or acute medication was prescribed, added, or changed as part of the study protocol.

Procedural adverse events (such as dizziness, worsening of pain, swelling, syncope, elevated blood pressure) were reported during and after the procedure (up to 60 min after).

The following baseline variables were collected for each patient: age, sex, duration of symptoms (in years), number and type of comorbidities (classified by system: cardiovascular, pulmonary, metabolic, rheumatological, neurological, gastrointestinal), and number of medications taken for headache management.

Clinical outcomes were assessed at baseline (T0) and at 1 month (T1). One month (4 weeks) was pre-specified as the primary assessment window, as it corresponds to the short-term efficacy interval conventionally used in greater occipital nerve block trials [14,15]; T1 therefore reflects the established window for evaluating the response to a single block. Outcomes assessed included:

• Pain intensity, measured on an 11-point NRS [16];

• Headache-related disability assessed using the 6-item Headache Impact Test (HIT-6). The HIT-6 is a tool used to measure the impact of headaches on the ability to function at work, at school, at home, or in social relationships [17].

• Sleep quality, evaluated with the Pittsburgh Sleep Quality Index (PSQI). The PSQI evaluates subjective sleep quality, sleep latency, sleep duration, usual sleep effectiveness, sleep disturbances, use of hypnotic drugs, and daytime disturbances in the previous month [18].

• Monthly headache frequency.

Each patient was instructed to keep detailed records of the number of headache crises using a diary that registered pain intensity, duration, preceding symptoms, triggers, and medication used. Patients were guided to anchor their estimates to the preceding 30 days, acute medication use, and routine activities to improve recall accuracy. For each patient, a crisis (attack) was defined according to the ICHD-3 criteria for that patient’s specific diagnosis (for example, a moderate-to-severe attack lasting 4–72 h for migraine and an attack of 15–180 min for cluster headache) rather than by a single uniform duration [19].

For each outcome variable, absolute change (Δ) from T0 to T1 was calculated as (T1 − T0). Percentage change (%Δ) was computed as %Δ = ((T1 − T0)/T0) × 100. With this convention, negative Δ and %Δ values indicate clinical improvement (lower NRS, HIT-6, PSQI scores, and fewer monthly crises).

We defined HIT-6 responders as patients achieving a ≥ 10-point reduction in HIT-6 score from T0 to T1. Prior work indicates that a ≥ 6-point decrease represents a clinically meaningful improvement in CM [20] and ≥ 8 points in chronic tension-type headache [21]. Therefore, we adopted a more conservative ≥ 10-point threshold—well above these minimal important change estimates—to enhance the specificity of the responder definition. This choice aims to (1) limit misclassification due to measurement variability and regression to the mean, (2) account for heterogeneity in minimal important change across headache phenotypes and baseline severities, and (3) ensure that “response” reflects a change of clear clinical salience in routine practice.

Descriptive statistics were used to summarize demographic and clinical data. Continuous variables are reported as mean ± standard deviation (SD) or median [interquartile range, IQR] as appropriate and compared using one-way analysis of variance (ANOVA) or the Kruskal-Wallis test when non-normally distributed. Categorical variables are expressed as frequencies and percentages and compared using Pearson’s chi-square test or Fisher’s exact test, as appropriate. Within-patient T0 → T1 changes were assessed with paired Wilcoxon signed-rank tests; reported Δ values and their 95% confidence intervals are Hodges-Lehmann estimates of the median paired difference and therefore need not equal the arithmetic difference between the two sample medians.

We performed univariate analyses to compare demographic characteristics, comorbidities, medication use, and clinical outcomes across the headache groups.

Subsequently, logistic regression models were constructed to explore associations between headache types and clinical variables (e.g., gender, medication use, and comorbidities). Linear regression models were applied to assess predictors of the T0 → T1 reduction in pain (NRS), disability (HIT-6), and sleep quality (PSQI) (absolute, in points) and in monthly crises (percent reduction). Variables with p < 0.10 in univariate comparisons were entered into multivariable models. Statistical significance was set at p < 0.05. To address (quasi-)complete separation in the response analyses, the multivariable model and the affected crude associations (cluster headache, rheumatic axis) were estimated using Firth penalized (bias-reduced) logistic regression, with profile-likelihood confidence intervals and penalized likelihood-ratio tests. All 29 patients had complete T0 and T1 outcome data (complete-case analyses); with 15 responders, the number of events per candidate predictor was low, which—together with quasi-complete separation—motivated penalized estimation and deliberately parsimonious models. The discriminative ability of baseline NRS for a HIT-6 response was assessed by receiver-operating characteristic (ROC) analysis, with the area under the curve (AUC) reported with a 5000-sample bootstrap 95% confidence interval and the optimal cutoff identified by the Youden index. As a sensitivity analysis addressing regression to the mean, the association between baseline NRS and the T0 → T1 change in HIT-6 was re-examined as a partial correlation adjusted for baseline HIT-6. Analyses were performed using MedCalc® for Windows version 20.217; Firth penalized logistic regression, the ROC analysis, and the partial-correlation sensitivity analysis were performed in Python (SciPy).

Results

A total of 29 patients were included, classified into four diagnostic categories: CM (n = 10), CH (n = 5), ON (n = 7), and CeH (n = 7). The mean age was 58.2 ± 13.7 years, with a female predominance (69%) (Table 1). A total of 23 procedures involved unilateral blocks.

Table 1.

Baseline demographics and comorbidity profiles across diagnostic groups

Variables	Chronic migraine	Cluster headache	Occipital neuralgia	Cervicogenic headache	p value
Number of patients (%)	10 (34.5%)	5 (17.24%)	7 (24.14%)	7 (24.14%)
Age (years), mean (± SD)	56.1 (± 16.04)	51.8 (± 7.82)	61.7 (± 9.58)	62.2 (± 16.54)	0.545
Sex (F/M)	7/3	2/3	6/1	5/2	0.407
Total comorbidities, mean (± SD)	1.6 (± 0.96)	2.40 (± 1.67)	0.86 (± 0.90)	1.71 (± 1.25)	0.250
Cardiovascular axis, n (%)	4 (40%)	3 (60%)	3 (43%)	4 (57%)	0.837
Pulmonary axis, n (%)	0 (0%)	1 (20%)	0 (0%)	1 (14%)	0.363
Metabolic axis, n (%)	3 (30%)	2 (40%)	2 (29%)	2 (29%)	0.652
Rheumatic axis, n (%)	4 (40%)	0 (0%)	0 (0%)	0 (0%)	0.032
Neurological axis, n (%)	4 (40%)	1 (20%)	0 (0%)	0 (0%)	0.082
Gastrointestinal axis, n (%)	1 (10%)	2 (40%)	1 (14%)	3 (43%)	0.322

Table 2 highlights between-group differences at baseline, specifically in medication patterns (acetaminophen, opioids, triptans, anti-calcitonin gene-related peptide (CGRP) monoclonal antibodies (mAbs), other antiepileptics) and in crisis frequency, while most other clinical scores were broadly comparable.

Table 2.

Baseline symptom duration, medication burden (overall and by class), and T0 clinical scores across diagnostic groups

Variables	Chronic migraine	Cluster headache	Occipital neuralgia	Cervicogenic headache	p value
Duration of symptoms (years), mean (± SD)	7.9 (± 2.9)	8.4 (± 3.6)	5 (± 3)	7.1 (± 2.7)	0.181
Number of medications, mean (± SD)	3.2 (± 0.79)	2.6 (± 0.89)	2.29 (± 0.49)	3.0 (± 0.58)	0.072
Of which, no. pts. (%)
Acetaminophen	4 (40%)	1 (20%)	4 (57%)	6 (86%)	0.048
NSAIDs	7 (70%)	2 (40%)	5 (71%)	2 (29%)	0.128
Corticosteroids	1 (10%)	1 (20%)	0 (0%)	3 (43%)	0.167
Opioids	0 (0%)	0 (0%)	1 (14%)	4 (57%)	0.012
Gabapentinoids	3 (30%)	3 (60%)	4 (57%)	4 (57%)	0.561
Antidepressant (TCAs, SNRIs)	3 (30%)	1 (20%)	2 (29%)	2 (29%)	0.981
Triptans	7 (70%)	5 (100%)	0 (0%)	0 (0%)	0.001
Anti-CGRP mAbs	3 (30%)	0 (0%)	0 (0%)	0 (0%)	0.033
Others antiepileptics	3 (30%)	0 (0%)	0 (0%)	0 (0%)	0.033
Number of crises per month T0, mean (± SD)	34.8 (± 15.08)	35.2 (± 8.68)	20.56 (± 5.36)	18.28 (± 2.12)	0.001
NRS T0, mean (± SD)	7.8 (± 0,91)	8.4 (± 0.89)	7.42 (± 0.53)	7 (± 1)	0.082
HIT-6 T0, mean (± SD)	66.3 (± 6,6)	68.4 (± 5.18)	63.2 (± 4.27)	64.7 (± 6.9)	0.524
PSQI T0, mean (± SD)	12 (± 2)	13.4 (± 3.28)	12 (± 2.76)	11.4 (± 3.15)	0.566
Bilateral block, n	2	0	3	1	0.316

CM chronic migraine, CH cluster headache, ON occipital neuralgia, CeH cervicogenic headache, NRS Numeric Rating Scale (0–10), HIT-6 Headache Impact Test-6, PSQI Pittsburgh Sleep Quality Index, TCAs tricyclic antidepressants, SNRIs serotonin-noradrenaline reuptake inhibitors, anti-CGRP mAbs anti-calcitonin gene-related peptide monoclonal antibodies, NSAIDs non-steroidal anti-inflammatory drugs, SD standard deviation

Across all measures, headache impact (HIT-6), sleep disturbance (PSQI), attack frequency, and pain intensity, paired Wilcoxon tests indicated consistent, clinically meaningful improvements, with Hodges-Lehmann confidence intervals favoring benefit and two-tailed p-values below the significance threshold. Specifically, median values significantly decreased from T0 to T1 for HIT-6 (66 vs. 52, p < 0.001), PSQI (12 vs. 8, p < 0.001), number of monthly crises (24 vs. 12, p < 0.001), and NRS pain score (8 vs. 4, p < 0.001), indicating clinically meaningful reductions in headache impact, sleep disturbances, attack frequency, and pain intensity (Table 3).

Table 3.

Wilcoxon scores

Score	Median T0	Median T1	Δ median	95% CI Δ	p value
HIT-6	66	52	− 12	[− 15, − 8]	< 0.001
PSQI	12	8	− 3.5	[− 5, − 2]	< 0.001
No. crises	24	12	− 14	[− 20, − 10]	< 0.001
NRS	8	4	− 3	[− 4, − 2.5]	< 0.001

Paired-sample Wilcoxon tests comparing baseline (T0) and post-treatment (T1) scores showed significant improvements across all measured outcomes; reported Δ and 95% CI are Hodges-Lehmann median paired differences, which may differ from the difference between the two sample medians

HIT-6 responders were 15 (51.7%), and they showed larger negative Δ across outcomes than non-responders. Overall, the responder/non-responder split cleanly separates T0 − T1 changes for pain, sleep, and attack burden (Supplementary Table S1).

Unadjusted analysis (Supplementary Table S2) showed longer symptom duration and greater baseline disease burden (higher monthly crisis counts and NRS scores at T0) in the HIT-6 responder cohort. Among medication classes, triptan use was more frequent in responders (higher odds of response), whereas anti-CGRP mAbs, opioids, antidepressants, and gabapentinoids showed no clear association. By diagnosis, CH appeared uniformly responsive in this small subgroup. Age, sex, overall medication count, and most comorbidity axes were not associated with response; the rheumatic axis was present only in non-responders, suggesting a possible negative association, though estimates were unstable. Variables with p < 0.10, together with prespecified covariates (age, sex, baseline HIT-6), were carried forward to the multivariable models.

A Firth penalized multivariable logistic regression model identified higher baseline NRS as a significant predictor of being a HIT-6 responder (OR = 3.38, 95% CI 1.22–13.47, p = 0.017). Symptom duration was not statistically associated with treatment efficacy (Table 4).

Table 4.

Firth penalized (bias-reduced) multivariable logistic regression for HIT-6 response (≥ 10-point decrease from T0 to T1), fitted to address quasi-complete separation

Variable	Std. error	p value	Odds ratio	95% CI for OR
NRS T0	0.61	0.017	3.38	1.22–13.47
Symptom duration (years)	0.15	0.253	1.18	0.88–1.58

Predictors retained in the final model: baseline pain intensity (NRS_T0) and symptom duration (years). Cells show standard error, p, odds ratio (OR), and 95% CI (OR per 1-point increase in NRS and per 1-year increase in duration). p values are penalized likelihood-ratio tests, and confidence intervals are profile penalized-likelihood intervals

NRS Numeric Rating Scale, PSQI Pittsburgh Sleep Quality Index

On receiver-operating characteristic analysis, baseline NRS discriminated HIT-6 responders from non-responders with an area under the curve of 0.82 (95% CI 0.65–0.95; Mann-Whitney p = 0.002). The Youden optimal cutoff was a baseline NRS ≥ 8, yielding 87% sensitivity and 71% specificity (positive predictive value 76%, negative predictive value 83%) for a clinically meaningful HIT-6 response (Fig. 2). Because predictive values depend on the responder prevalence in this sample (51.7%), the reported positive and negative predictive values may not transport to settings with different base rates. Given the small sample and the single-center design, this threshold and its diagnostic metrics (sensitivity, specificity, and predictive values) should be regarded as hypothesis-generating and warrant prospective validation.

Fig. 2.

Receiver-operating characteristic (ROC) curve for baseline pain intensity (Numeric Rating Scale, NRS) as a predictor of a clinically meaningful Headache Impact Test (HIT-6) response (≥ 10-point reduction at one month) in 29 patients. Area under the curve = 0.82 (95% CI 0.65–0.95). The Youden optimal cutoff (baseline NRS ≥ 8; red square) gave 87% sensitivity and 71% specificity. The dashed line denotes chance

In a sensitivity analysis addressing regression to the mean, higher baseline NRS remained associated with a greater T0 → T1 reduction in HIT-6 after adjustment for baseline HIT-6 (partial r = 0.48, p = 0.011).

Separate models were fitted for the T0 → T1 reduction in NRS, HIT-6, and PSQI (absolute, in points) and in monthly crises (percentage). Across models, two signals were consistent: patients with CH showed greater improvement, whereas a rheumatic-axis comorbidity was associated with less improvement. As anticipated on statistical grounds, within-domain baseline severity was associated with larger within-domain reductions (higher baseline NRS, HIT-6, and PSQI each predicting a larger reduction on the same scale); because the change score incorporates the baseline value, these within-scale associations are expected from mathematical coupling and regression to the mean and are not interpreted as clinical predictors. In contrast, age, sex, total comorbidities, symptom duration, and baseline attack frequency were not retained as independent predictors after adjustment. Supplementary Tables S2–S5 report coefficients (β), standard errors, partial/semipartial correlations, and p-values for each outcome, together with model diagnostics; assumption checks supported residual normality, homoscedasticity, and linearity.

Discussion

In this mixed‑diagnosis cohort of 29 patients treated with a GONB targeted at C2, a large, clinically meaningful short‑term improvement was observed at T1 (1 month): median NRS pain fell from 8 to 4, median HIT‑6 from 66 to 52, median PSQI from 12 to 8, and median monthly crises from 24 to 12 (all p ≤ 0.001). Higher baseline pain intensity (NRS_T0) independently predicted achieving a clinically meaningful HIT‑6 response (≥ 10‑point reduction; OR 3.38, 95% CI 1.22–13.47, p = 0.017). Multivariable modeling showed diagnosis‑dependent effects (CH associated with larger improvements across NRS, HIT‑6, and crises) and that rheumatic comorbidity predicted smaller treatment gains across multiple outcomes. These results indicate that a single, ultrasound-guided proximal (C2) GONB is associated with rapid and substantial short-term improvements across several headache phenotypes and that baseline clinical features and comorbidity profiles help explain inter-individual variability in response.

Technique and Procedural Choices

The C2 approach, with in-plane lateral-to-medial injection and confirmation by perineural fluid distension, aligns with the recent movement in the literature toward image-guided proximal targeting for greater anatomical precision and reproducibility. Previous prospective work validating an ultrasound C2 approach reported high early block success (~ 86% at 30 min) and clinically meaningful NRS reductions that, although attenuating over time, remained significant through 4 weeks [5]. More recent prospective work using the same ultrasound-guided proximal C2 technique reported comparable improvements in pain and HIT-6 in chronic migraine, supporting the reproducibility of this approach [22].

Ultrasound guidance and proximal targeting can increase accuracy and are associated with larger or more durable reductions in pain and headache days compared with blind landmark approaches, especially when measured up to ~4 weeks after a single injection [23]. In fact, the use of ultrasound to visualize the nerve between semispinalis and obliquus capitis inferior at C2 improves anatomic specificity and avoids some limitations of superior‑nuchal‑line landmark techniques (e.g., bony artifacts, distal branching), a rationale that relies on the fact that “visualizing the nerve and then performing the injection is more accurate than detecting the artery with Doppler ultrasonography and performing the injection just medial to the artery” [23]. Furthermore, this technique reduces the risk of deep misplacement and inadvertent intracranial spread; however, attention to structural contraindications and conservative injectate selection remains essential to prevent major complications [11].

A unilateral approach was preferred in almost all blocks performed because pain was reported as unilateral in most of the enrolled patients. Moreover, several trials reported that the use of an ipsilateral approach is efficacious in reducing pain in several cases of secondary headaches and migraine [24]. Finally, a unilateral approach ensures a faster, more comfortable procedure for patients who cannot tolerate prone positioning for an extended period.

A combination of LA with low doses of particulate steroid was chosen to enhance the block effect, despite the ongoing debate regarding the use of steroids [25]. Current guidelines for migraine prevention provide a weak recommendation against the use of corticosteroids [26]. On the contrary, corticosteroids have been used in peripheral blocks for CH [24], despite reported adverse events, both local and systemic [27]. The low dose of dilute particulate steroid and direct echographic visualization during injection ensure the safety of this block, as demonstrated by the absence of adverse events in the present study.

Response across Headache Phenotypes

Regarding the diagnosis-specific effects, our results showing that CH predicts larger improvements across outcomes (ΔNRS, ΔHIT‑6, ΔPSQI, Δcrises) are consistent with clinical reports and data indicating that interventions targeting the occipital region often produce pronounced but typically transient benefits in CH populations [4,25,28]. This supports the hypothesis that peripheral occipital blockade can interrupt pathways particularly relevant to cluster pathophysiology or its downstream expression (e.g., trigeminovascular and autonomic circuits linked to the hypothalamus) [2]. This phenotype-specific benefit aligns with current European guidance, which recommends GONB as a treatment option in cluster headache [29].

A similar but less pronounced trend was seen in CM, ON, and CeH, where each subgroup showed reductions in NRS and HIT‑6 at 1 month, which was consistent in direction and magnitude with prior studies [4,15,23,30–33] and consistent with GONB as a viable short‑term adjunct/bridge therapy in selected patients belonging to these subgroups. Quantitatively, the pain relief observed in our cohort (median NRS 8 → 4) is consistant with controlled evidence: a multicenter randomized trial reported a comparable 1-month reduction in pain after greater occipital nerve blockade (VAS 8.4 → 5.3, achieved with four weekly blocks) [15], and a meta-analysis of six randomized trials confirmed significant reductions in pain score and headache days versus placebo [14], supporting the external validity of our real-world findings despite the uncontrolled design. Notably, the pooled controlled effect on pain in that meta-analysis was modest (standardized mean difference ≈ − 0.5), so the larger uncontrolled reduction observed here should be interpreted with the caveats noted below.

Predictors of Response and Possible Mechanism

Our data suggest that patient selection critically shapes the clinical outcome after GONB. In adjusted analyses, higher baseline pain intensity was an independent predictor of a clinically meaningful improvement in headache impact at 1 month. Practically, this means that patients who present with more severe pain are more likely to perceive a large benefit after GONB. This finding can therefore be considered when counseling candidates about the procedure. We also observed phenotype- and comorbidity-specific modifiers: patients with CH tended to experience greater benefits, whereas patients with rheumatic comorbidities showed attenuated improvements across pain, impact, and attack frequency. These patterns, rarely explored in previous GONB literature, which often focuses on single diagnoses (such as CM [34] or CH [25]) and rarely profiles systemic comorbidities, point to a possible biological mechanism (e.g., systemic inflammatory tone and central sensitization) that could dampen the relative effect of a focal peripheral block.

Taken together, baseline pain intensity was the most useful predictor of response, with a baseline NRS ≥ 8 best separating responders from non-responders (Fig. 2). The phenotype and comorbidity signals pointed in the same direction: all five cluster patients responded, whereas none of the four with a rheumatic-axis comorbidity did, though these subgroups were too small to define firm thresholds. We therefore offer baseline pain, cluster phenotype, and rheumatic comorbidity as hypotheses for patient selection to be confirmed prospectively rather than applied as a decision rule.

Mechanistically, these observations are consistent with the convergence of occipital and trigeminal afferents onto second-order neurons of the trigeminocervical complex, an anatomical and physiological substrate demonstrated in animal models that underlies input-induced central sensitization and the referral of pain across cranial territories [35,36]. By reducing nociceptive drive from greater occipital nerve afferents into this shared relay, a proximal block may attenuate central sensitization rather than act only on a local pain generator, consistent with its cross-phenotype efficacy and with its documented modulation of trigeminocervical nociceptive processing [35]. High baseline pain intensity may indicate greater central sensitization and, therefore, greater potential to benefit from a reduction in afferent input; conversely, a systemic inflammatory or rheumatic background may sustain central sensitization through mechanisms that a focal peripheral block does not address. This offers a coherent rationale for the divergent responses we observed.

Effects on Sleep and Clinical Relevance

Importantly, symptomatic improvement was not limited to pain and headache. An upgrade in sleep quality over the first month supports the bidirectional relationship between sleep and headache and is consistent with the idea that reducing peripheral drive to trigeminocervical pathways can secondarily normalize arousal networks [2].

The median HIT‑6 reduction in our cohort (− 12 points) exceeds commonly cited minimal detectable/clinically important thresholds and is larger than the smallest detectable change (SDC) reported in psychometric validation work (e.g., HIT‑6 SDC ∼4.38 in one validation [17]), indicating that the observed change is both statistically and clinically meaningful. Similarly, a 50% reduction in monthly crises and a 3‑point median NRS drop represent clinically relevant improvements for most patients and compare favorably with effects reported in many previous short‑term GONB series and RCTs [6,15,31–33,37]. That said, the evidence is not uniform: at least one randomized, placebo-controlled trial found no benefit of occipital nerve block over placebo for migraine, which reinforces the need to interpret uncontrolled improvements such as ours with caution [38]. Accordingly, these uncontrolled before-after findings should be interpreted as associations and weighed against this controlled evidence.

Strengths and Limitations

Therefore, the novelties of this study are: (1) clinically meaningful 1-month improvements in pain, headache impact, sleep, and attack frequency after a single ultrasound-guided proximal (C2) GONB in a mixed real-world cohort including patients with CM, CH, CeH, and ON.

(2) Identification of baseline pain intensity as an actionable predictor of HIT-6 response, and proportionally greater benefit in CH.

(3) A new comorbidity signal: rheumatic axis associated with attenuated improvement, suggesting systemic factors modulate GONB effectiveness.

This study presents several limitations. First, the relatively small sample size, especially within individual headache subtypes, may reduce the statistical power of subgroup comparisons and limit the generalizability of the findings. The small sample reflects a single-center series restricted to adults with ICHD-3-confirmed chronic, drug-refractory headache, which limited the number of eligible patients despite the 1-year enrolment period and the spread across four diagnoses. Second, the assessment was pre-specified at 1 month—the short-term window conventionally used to evaluate a single GONB [14,15]—so these data characterize short-term efficacy by design; durability of benefit beyond this window was not assessed because real-world attrition after T1 was substantial and likely non-random (patients not returning either because they had improved or because they had disengaged), precluding an unbiased longer-term analysis. Third, the absence of a control group prevents definitive attribution of observed effects to the intervention. Fourth, the use of patient-reported outcome measures, while clinically relevant, may be influenced by subjective perception and reporting bias. Fifth, several different diagnoses (CM, CH, CeH, ON) were considered, introducing another bias in both patient selection and data analysis. In particular, monthly attack counts were pooled across phenotypes with differing attack definitions, a limitation of the frequency outcome. Sixth, part of the observed relationship between higher baseline pain and greater improvement may reflect regression to the mean and floor effects rather than a genuine treatment effect; we mitigated this by measuring response on the HIT-6, an instrument distinct from the NRS predictor, using a conservative ≥ 10-point responder threshold, and, in a sensitivity analysis, higher baseline NRS remained associated with HIT-6 improvement after adjustment for baseline HIT-6.

Lastly, the lack of correction for multiple comparisons may increase the risk of a type I error. These limitations underscore the need for future randomized controlled trials with larger sample sizes and longer follow-up periods to validate and extend our findings.

Conclusions

In this prospective real-world cohort, a single ultrasound-guided proximal (C2) GONB was followed by rapid, clinically meaningful improvements at 1 month in pain, headache impact, sleep, and monthly attacks, across chronic migraine, cluster headache, occipital neuralgia, and cervicogenic headache, and with no adverse events. The procedure is safe, quick, and inexpensive, has few drug interactions, and fits naturally as a bridge or adjunct for patients who have run out of pharmacological options. Its benefit is also predictable rather than random: it was largest in patients with more intense baseline pain and in the cluster phenotype, which provides the clinician with usable criteria for selecting candidates and counseling them before the block. These features make proximal ultrasound-guided GONB a practical, patient-tailored option in the management of refractory chronic headache.

Supplementary Information

Below is the link to the electronic supplementary material.

Author Contributions

Alberto Corriero and Mariateresa Giglio conceived and designed the study and wrote the first draft of the manuscript; Alberto Corriero also performed all the statistical analyses and prepared the figures and tables. Carlo Tramacere collected the clinical data and contributed to data interpretation. Angela Preziosa, Fara Fornarelli, Filomena Galdini, Alfonso Pilolla, and Salvatore Grasso contributed to the acquisition and interpretation of data. Giustino Varrassi and Antonella Paladini provided expert clinical input and contributed to the interpretation of data. Filomena Puntillo supervised the study as senior author and contributed to the study design and data interpretation. All authors critically revised the manuscript for important intellectual content, approved the final version to be published, and agree to be accountable for all aspects of the work.

Funding

No funding or sponsorship was received for this study or publication of this article.

Data Availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Declarations

Conflict of Interest

Alberto Corriero, Mariateresa Giglio, Carlo Tramacere, Angela Preziosa, Fara Fornarelli, Filomena Galdini, Alfonso Pilolla, and Salvatore Grasso have nothing to disclose. Giustino Varrassi, Antonella Paladini, and Filomena Puntillo are Editorial Board members of Pain and Therapy. They were not involved in the selection of peer reviewers for this manuscript nor in any of the subsequent editorial decisions.

Ethical Approval

This study was performed in accordance with the ethical principles of the Declaration of Helsinki of 1964 and its later amendments and with the International Association for the Study of Pain (IASP) guidelines for pain research in humans. The study protocol was approved by the local Ethics Committee (Comitato Etico Locale “Gabriella Serio,” IRCCS Istituto Tumori Giovanni Paolo II, Bari, Italy; protocol/IRB no. C2832/CEL—STUDIO NGO 01). Written informed consent to participate was obtained from all individual participants. No identifiable individual patient data are reported; therefore, consent for publication was not applicable.