Deep Brain Stimulation for Chronic Cluster Headaches: A Systematic Review and Meta-Analysis Free

Cluster headaches are severe, unilateral, periorbital headaches that may last 15–180 min and occur in clusters. It is one of the most severe pain syndromes, sometimes referred to as a “suicide headache” due to increased suicidal thoughts during attacks [1]. Management of cluster headaches includes (1) preventative medications, such as verapamil and corticosteroids, (2) acute therapies, such as oxygen and injectable sumatriptans, and (3) neuromodulation techniques in severe and resistant cases including noninvasive vagus nerve stimulation, occipital nerve stimulation (ONS), and sphenopalatine ganglion stimulation (SPGS) [2]. Unfortunately, in about 10–15% of cases, patients develop chronic cluster headaches (CCHs). Consequently, the diagnosis of CCH is given to patients who experience cluster headaches without a remission period for at least 1 year, or with remissions lasting <3 months [3].

Therapeutic deep brain stimulation (DBS) has been applied mainly for constant neuropathic pain syndromes and not for intermittent episodic chronic pain syndromes [4]. Based on neuroimaging findings of posterior hypothalamic activation during cluster attacks [5], Leone and colleagues from Milan, Italy published the first case of DBS for the treatment of CCH in 2001 where they targeted the posterior hypothalamus [6]. Since then, this neuromodulation technique has been reported in more than 100 cases using various anatomical targets in the vicinity of posterior hypothalamus. The aim of this study was to perform a comprehensive systematic literature review and meta-analysis of patients with CCHs who have been treated with DBS to provide insight on safety and surgical outcomes.

A systematic review and meta-analysis were performed according to the PRISMA 2020 guidelines [7]. The searched databases included PubMed and Scopus. The search term was deep brain stimulation and headache*. The search was conducted on January 3, 2022, and all studies published to that date were reviewed. The reference lists of included studies were also reviewed. Duplicates were removed manually within Microsoft Excel.

Each study was independently reviewed by the two authors (M.M. and A.A.). The screening process of studies involved two consecutive steps: first, by title and abstract, and second, by full manuscript. Included studies were clinical series that used DBS for CCH and that reported outcomes including post-treatment headache attack frequency and/or intensity. Exclusion criteria included animal studies, studies written in a language other than English, other neuromodulation techniques for CCH, and studies with insufficient outcome data or irrelevant endpoints. Abstracts, review articles, textbook chapters, comments, and single-patient case-reports were excluded. For reports with extensive overlapping patients at the same institution, the publication with the larger cohort was included. When there was not significant overlap, all studies were included, and duplicate patients were removed from analysis. The authors used the Newcastle-Ottawa Scale (NOS) [8] for assessing the quality of included studies.

Data extraction was performed manually and included basic study information: title, authors, study design, date of publication; patient demographic data: age, gender; headache information: duration of disease, laterality of headache, headache attack frequency before and after DBS, headache intensity scores before and after DBS, quality of life (QoL) before and after DBS, and follow-up; DBS details: manufacturer, anatomical target, type of anesthesia, use of microelectrode recording, laterality of lead placement; and surgical complications. To standardize the frequency of headaches between studies, various units were converted to headaches per week. Across all studies that reported headache intensity, intensity was measured in a standardized fashion from 0 to 10, with 0 signifying no pain and 10 describing the most intense pain.

Descriptive statistics and pooled-analysis were performed using IBM SPSS (IBM Corp., Armonk, NY, USA). Meta-analysis was performed according to PRISMA 2020 guidelines. Review Manager (RevMan. Version 5.4. The Cochrane Collaboration) was used to perform meta-analysis. Mean difference (MD) was used for continuous variables (e.g., headache attack frequency per week, headache intensity). Heterogeneity was quantified using I² where I² > 50% was considered to be moderate to high heterogeneity. The random-effects model was used for analysis to account for heterogeneity between studies. Two-tailed t test, Wilcoxon signed-rank test, and Mann-Whitney U test (Wilcoxon rank-sum test) were calculated to evaluate for statistical significance. All statistical tests used an alpha level of 0.05 or less and a 95% confidence interval to test for statistical significance. For purposes of data analysis, responders to DBS were defined by ≥50% improvement in headache frequency or intensity, and nonresponders were defined by <50% improvement.

The initial search resulted in 797 studies: 282 from PubMed and 515 from Scopus. After 237 duplicates were removed, 560 studies were screened. Of those, 486 studies were excluded and 79 were sought for full manuscript retrieval. Additional 5 studies were included in full manuscript retrieval from cross-references. Sixty-three studies were excluded after investigation of the full manuscript. Subsequently, 16 final studies were included in the review [9‒24]. A summary of this workflow is shown in a PRISMA flow diagram in Figure 1. Based on the NOS for assessing the quality of studies, all reports were graded as “good” quality based on the grading system of good, fair, or poor.

Of these 16 studies, 12 studies described unique patient cases and 4 studies [9, 11, 12, 16] provided supplementary details to those cases. The designs of these 12 studies were 1 double-blind randomized controlled trial (RCT), 9 prospective trials, and 2 retrospective trials. Three patients with no clinical or follow-up in an eight-patient series were excluded [15]. One case was aborted due to intraoperative panic attack [10] and was excluded from further outcomes analysis.

A total of 108 unique patient cases of CCHs treated with DBS were identified. Patients ranged from 24 to 71 years old with an average age of 46.6. Men comprised 76.9% (n = 83) of patients. Unilateral headaches were experienced in 86.3% (n = 88) of patients. Of those with unilateral headaches, 60.2% (n = 53) experienced left-sided headaches. At baseline, the average (SD) headache attack frequency per week was 33.8 (16.7), and the average headache intensity score, on a scale from 0 to 10, was 8.5 (1.8). On average, duration of disease from diagnosis to DBS treatment was 9.5 years (6.7), ranging from 1 year to 35 years. Further clinical characteristics are described in Table 1.

The operation was performed under local anesthesia with minimal sedation in 82.4% (n = 89) of patients and under general anesthesia in 17.6% (n = 19) of patients. Intracranial leads were placed unilaterally in 82.4% (n = 89) of patients and bilaterally in 17.6% (n = 19). Of the 101 cases that reported the DBS system used, 95% (n = 96) patients were implanted with leads by the same manufacturer. The anatomical target for intracranial lead placement was the posterior hypothalamus in 58.3% (n = 63) of cases, ventral tegmental area in 21.3% (n = 23), pre-rubral tegmentum and mammillothalamic tract in 13.9% (n = 15), and posterior inferior third ventricle floor in 6.5% (n = 7). Microelectrode recording (MER) was used in 40.8% (n = 29) of the 71 cases that commented on whether MER was used or not [10, 13, 15, 22]. All 29 cases that used MER were performed awake and none of the cases that underwent general anesthesia has utilized MER. The average overall follow-up period was 45.4 months (38.7) and ranged from 1 to 144 months. All patients had at least 6 months of follow-up except 1 patient whose implant was removed at 1 month due to infection.

The average headache attack frequency per week decreased from 33.8 to 10.1 (70.1% decrease) after DBS. The average headache intensity score decreased from 8.5 to 4.2 (50.6% decrease) after DBS (Figure 2a, b). Out of the 105 cases that reported outcomes, 78 (74.3%) patients were considered responders based on ≥ 50% improvement in headache frequency. In pooled-analysis, the difference in headache attack frequency and headache intensity compared between before and after DBS were statistically significant with a p value <0.001 for both comparisons. Meta-analysis revealed that the MD in headache attack frequency before and after DBS (MD: 20.97, 95% CI: 13.58–27.01, p = 0.0001) and in headache intensity before and after DBS (MD: 5.00, 95% CI: 3.16–6.84, p < 0.0001) were statistically significant. Forest plots regarding these two metrics are shown in Figure 3a, b.

The use of microelectrode recordings during implantation significantly impacted headache intensity outcomes. There was no difference in baseline headache frequency or intensity between patients who received MER and those who did not. Utilization of intraoperative MER during lead placement was associated with statistically significant improvement in headache intensity postoperatively (p = 0.006), favoring the use of MER. The average (SD) postoperative headache intensity score was 3.05 (2.53) when MER was used as opposed to 5.18 (3.17) without MER. Similarly, the use of MER improved postoperative headache frequency; the average frequency was 6.05 headaches/week with MER as opposed to 12.17 headaches/week without MER. However, this was not shown to be statistically significant. There were no significant differences in outcomes associated with age, gender, or awake versus asleep implantation.

All studies included reported complications. These were due to placement, device, stimulation amplitude, and surgery-related superficial and deep surgical-site infections (SSI). Major complications included electrode misplacement or breakage (7.41%, n = 8), SSI requiring additional surgery (4.63%, n = 5), neurological deficits (3.70%, n = 4), contralateral (side-shift) cluster attacks (2.8%, n = 3), and intracerebral hemorrhage (0.93%, n = 1). Death occurred in 1 patient, who suffered an intracerebral hemorrhage. Minor complications included SSI which resolved with antibiotics alone (2.78%, n = 3), asymptomatic hemorrhage (0.93%, n = 1), and micturition syncope (0.93%, n = 1). Further detail regarding complications with the corresponding interventions and outcomes is provided in Table 2. Most common transient complications included 24 cases (22.2%) of diplopia, 19 cases (17.6%) with changes in satiety/hunger, sexual drive/hormones, and 7 cases (6.5%) of vertigo or dizziness. These were related to stimulation amplitude and most of them were resolved with lower intensity.

DBS is a feasible surgical technique that can be successfully performed either under local or general anesthesia for patients suffering from CCHs who have had insufficient headache control with medical management. In addition, many of these patients have previously failed other neuromodulation techniques such ONS or SPGS. This further emphasizes the necessity for an additional treatment option for managing this devastating pain syndrome. In all patients included in this review, there was only one case aborted due to intolerance (<1%), but no cases were aborted due to technical challenges. Since only 5 patients out of the 19 asleep cases had individual data, direct comparison of awake versus asleep could not be performed.

It also appears that DBS for CCH has a reasonable safety profile and is well-tolerated by patients. However, as with all surgical interventions, DBS has associated risks. Fortunately, based on this review, the mortality rate was <1% as there was only one death out of the 108 patients. On the other hand, the rate of major complications was 16.67%. Even though this number may seem elevated, it is comparable with the complication rates of DBS used for other indications [25‒28]. Moreover, it is important to note that posterior hypothalamus/ventral tegmentum is a relatively novel target, unfamiliar to even experienced neurosurgeons. As this treatment modality continues to advance, and neurosurgeons become more accustomed with this target, better outcomes with less frequent complications will likely result in the future.

This is the most complete systematic review and meta-analysis on this topic to date. DBS for CCH appears to be an effective neuromodulation therapeutic option for resistant cluster headache. In this meta-analysis, both headache attack frequency and intensity were significantly decreased with DBS. Headache control for the patients within this review seems very effective with almost 70% of patients achieving excellent results. An additional 10% of patients had moderate improvement in their headaches with DBS. Despite the vast majority of studies reported significant improvement in CCH following DBS, the RCT study conducted by Fontaine and colleagues showed no improvement [14]. The authors found no significant difference in outcomes among the 11 patients who were randomized for the stimulation versus sham. Subsequently, DBS was considered “probably ineffective” by the American Headache Society [29]. However, this study has several limitations including the small sample size and the short randomization period of 1 month only. Moreover, at the end of the open phase, 6 patients reported significant improvement and refused to turn off the stimulator in order to participate in a new randomization phase.

It is also worth mentioning that other neuromodulation techniques have evolved such as noninvasive vagus nerve stimulation, ONS, and SPGS [2]. Due to their less invasive nature and good outcomes, these techniques became popular to a level that some neurosurgeons, who previously published on DBS for CCH, started to consider them before DBS [30, 31]. Nevertheless, these modalities have their specific complications and limitations. For example, a large percentage (up to 36%) of patients treated with ONS has developed contralateral CCH [32], compared to approximately 3% in DBS based on this systematic review. Actually, some authors proposed that DBS might protect against developing this side-shift phenomenon [31]. Moreover, several reports in this systematic review had a large percentage or all their patients previously failed such less invasive techniques before improving later following DBS surgery [15, 17, 19, 24]. Hence, careful selection of patients for DBS is crucial.

The impact on patient QoL is of paramount importance. Prior to surgery, patients in the Broggi et al. [12] cohort were said to be in “poor condition.” 1 patient had such intense pain that he had attempted suicide twice, and others had severe side effects from chronic medication use (steroids, triptans). These patients had consistently improved QoL postoperatively. Both Akram and Nowacki groups have assessed the QoL using the 36-item short form survey (SF-36) and reported a statistically significant improvement in the physical [19, 23] and mental components [23]. Other groups used the hospital anxiety and depression (HAD) scale where patients reported a statistically significant decrease in anxiety [14, 20] and depression [20]. After treatment with DBS, all patients treated by Seijo et al. [22] could return to work, and all patients treated by Piacentino [18] reported improved family relationships. This demonstrates how limiting untreated CCH can be, and how effective DBS is for intractable CCH. Due to variability in tools used to assess QoL across the studies, direct comparison was not feasible.

Since the first case of DBS for CCH by Leone et al. [6] in 2001, several different surgical techniques have been described trying to localize the ideal target. Initial studies focused on targeting the posterior hypothalamus grey matter based on a landmark imaging study that showed ipsilateral posterior inferior hypothalamic activation during a cluster attack that was not seen when patients were asymptomatic [5]. However, subsequent studies have questioned the posterior hypothalamus and suggested that the best target is actually further “posterior” to the posterior hypothalamus per se and involves the diencephalon-mesencephalon junction, mainly the ventral tegmentum including mammillotegmental tract, substantia nigra, and red nucleus [33]. They attributed this difference due to less precision of PET scans to localize very small regions in the brain and variations in human atlases. Based on these observations, several neurosurgery groups have adjusted the original coordinates by the Milan group (X = 2 mm, Y = −6 mm, Z = −8 mm) [6]. The group themselves evolved their technique and changed coordinates over the years to (X = 2 mm, Y = −3 mm, Z = −5 mm) [9]. Both targets were used in most of the published studies to date. Interestingly, studies that analyzed the active lead contacts found that they were actually located outside the hypothalamus and more posterior to the intended target [23, 34]. Figure 4 summarizes the targets based on the various coordinates used across the studies.

Due to the quality of the published data, comparing outcomes of posterior hypothalamus with ventral tegmental area as targets was not possible. Not all reports have assessed the location of effective contact points postoperatively. In addition, in the studies that reported these coordinates, they varied across patients within the same study. Nowacki et al. [35] have meta-analysed individual data from 40 published cases where they identified two hotspots of stimulation including the ventral tegmentum and retrorubral area. Nevertheless, our meta-analysis demonstrated the positive impact of DBS on CCH regardless of the intended target in this hub. Moreover, device technology has dramatically advanced since most of these studies. Several companies have developed directional DBS leads which have been recently approved by the FDA. Only one study [24] reported that directional leads were used in “some patients” without stating patients who actually had them or the company of the leads used. Since the posterior border of hypothalamus is close to the ventral tegmentum, having this new technology of directional leads will enable the treating providers better select contacts for optimal areas for stimulation.

The effect of utilizing MER for targeting on headache outcomes is of interest. This technique involves inserting a small electrode into the intended target and record spontaneous and induced neural activity including single-neuron activity and local field potential. MER helps refine final target placement of the DBS lead by adjusting the electrode position based on the specific firing pattern of nucleus of interest. In this meta-analysis, the use of microrecording was significantly associated with even better headache intensity scores. In contrast to the use of MER for targeting in the subthalamic nucleus, globus pallidus or ventralis intermediate nucleus of the thalamus, specific neuronal activity of the posterior hypothalamus or ventral tegmentum area has only been partially characterized. It appears that spontaneous discharges in this region are characterized by low frequency (12–24 Hz) compared to higher frequency of DBS targets in movemen disorders [36, 37]. In the included studies that reported the stimulation parameters, most of them used frequency of 180 Hz, pulse width of 60 μs, and amplitude of 2 V.

Nonetheless, target was not adjusted based on MER in any of the studies that utilized this technique [10, 13, 15, 22]. Hence, the difference in outcomes may be due to other factors. It is also worth noting that MER may increase the risk of bleeding as the only fatality in this meta-analysis was due to catastrophic bleeding after MER [10].

Cluster headaches are considered the most severe of primary headaches. The pain is excruciating with associated considerable personal burden due to lifestyle restrictions, reduced productive capacity, and increased use of healthcare [38]. Ford and colleagues [39] published a socioeconomics study in 2017 to estimate the cluster headache cost burden in the USA where the total direct and indirect costs were $8,060 per insured patient per year based on the US dollar value in 2015. Patients with CCH unfortunately have further burden; their healthcare cost is 5 times higher than patients with episodic CH bouts with an estimated cost of €13,350 based on the Euro value in 2019 [40]. Furthermore, approximately 10–20% of patients lose their job secondary to cluster headache [38, 40]. Hence, more effective treatment is warranted to decrease burden on patients and the healthcare system.

In an attempt to evaluate the cost of DBS therapy for CCH, Leone and colleagues conducted a retrospective cost-effectiveness study comparing the cost of DBS with sumatriptan injection at baseline before surgery [41]. Interestingly, DBS was more cost-effective with approximately €15,000 of savings per patient over a period of 8 years based on the Euro value in 2008. In another study by Arkam et al. [19], the triptan intake in their series decreased by approximately 60% with a yearly saving of £4,750 per patient on triptans alone.

Although it has been over 20 years since the first study of DBS for CCH, only one pilot RCT has been conducted [14] and no publicly registered RCTs are ongoing currently. Hence, more robust clinical trials are needed. Furthermore, the advent of directional leads could potentially further improve outcomes and decrease associated complications. In addition, different patterns and waveforms of brain stimulation may further improve outcomes as has been well described for spinal cord stimulation for many different pain phenotypes. All these factors could help justify this therapy with FDA approval with perhaps even a Humanitarian Device Exemption.

This study has several limitations. First, the sample size of 108, out of 16 studies, is not large and eventually has limited sub-groups analysis. There were no robust clinical trials in the studies included. Most were prospective cohorts without blinding or control subjects. Furthermore, the studies were heterogenous necessitating conversion of outcome units to make comparison between studies possible. Excluding non-English papers is also a limitation considering that some studies of DBS for CCH were completed or ongoing by groups not publishing in English language journal.

DBS for CCHs is a feasible surgical technique with a reasonable safety profile that can be successfully performed either awake or asleep. In carefully selected patients, especially, when previously failed minimally invasive neuromodulation techniques, approximately 70% of patients achieve excellent control of their headaches with DBS. Utilization of microelectrode recordings maybe associated with better outcomes.

An ethics statement is not applicable because this study is based exclusively on published literature.

The authors have no conflicts of interest to declare.

This project did not require funding support.

Murray: data collection and manuscript drafting. Pahapill: critical revision of the work and interpretation. Awad: conception and study design, data collection, and manuscript drafting.

Worksheet data are not available for public access due to legal purposes and copyright of journals where data were extracted from. Further inquiries can be directed to the corresponding author.