Factors associated with efficacy of occipital nerve stimulation in medically intractable chronic cluster headache

Abstract

Background/Objective

Occipital nerve stimulation (ONS) has become an established therapy for medically intractable chronic cluster headache (MICCH), but unfortunately, one third of the patients do not respond satisfactorily. Reliable predictors of treatment success would help physicians improve indication for ONS in MICCH. Although a recent report suggested several factors that were associated with treatment failure (early onset of cluster headache [CH], chronic cluster headache [CCH], and smoking) this study was small, did not use a formal model, efficacy was poorly defined, and the follow‐up was only of short duration. Here, we retrospectively sought: (i) reproduction of these associations and (ii) identification of possible other associations in our previously published double‐blind randomized controlled “Occipital Nerve Stimulation in Medically Intractable Chronic Cluster Headache” (ICON) trial, and long‐term follow‐up, of the efficacy of ONS in MICCH.

Methods

Data from the double‐blind randomized controlled ICON trial, and its prospective open‐label extension, were analyzed in this prospective cohort study in the Netherlands (October 12, 2010, to December 20, 2020) for: (i) relative differences in attack frequency and (ii) subjective satisfaction with effect between baseline and at 4 and 24 weeks, and 2 and 5 years, after ONS implantation. Formal statistical models were used to: (i) verify the previously detected associations and (ii) identify possible other associations.

Results

Early onset of CH and smoking did not predict efficacy of ONS. Relative reduction in attack frequency at 24 weeks (B = 0.44, 95% confidence interval [CI] 0.13–0.76; p = 0.007) and the time since onset of CCH (B = 4.04, 95% CI 1.16–6.92; p = 0.007) appeared to be the only factors that were associated with objective efficacy at 2 years, and relative attack reduction after 2 years was the only factor associated with objective efficacy at 5 years (B = 0.501, 95% CI 0.186–0.815; p = 0.003). The odds of experiencing subjective satisfaction with ONS after 2 years increased with a later debut of CCH (adjusted odds ratio [aOR] 1.06, 95% CI 1.01–1.12; p = 0.033) and greater relative reduction in attack frequency at 24 weeks (aOR 1.02, 95% CI 1.00–1.04; p = 0.017).

Conclusion

In a controlled setting, early onset of CH, CCH, and smoking were not associated with treatment success of ONS for MICCH, as previously suggested by others in an uncontrolled setting. Early response at 24 weeks after initiation of ONS was the only factor that was associated with long‐term efficacy, which was identified. Since a large proportion of patients with MICCH improve with ONS, we recommend offering ONS to all patients with MICCH.

Plain Language Summary

Previous research has not identified any reliable predictors of treatment response related to the use of occipital nerve stimulation (ONS), but early onset of (chronic) cluster headache and smoking may be associated with treatment failure. We investigated data from a randomized controlled trial and its long‐term follow‐up of occipital nerve stimulation (the ICON trial), and did not find any association between these suggested predictors and treatment response. Our results only showed an association between a positive treatment response at 24 weeks and a positive long‐term treatment response, and since a large proportion of patients with medically intractable chronic cluster headache (MICCH) improve with ONS, we recommend offering ONS to all patients with MICCH.

Abbreviations

CCH

chronic cluster headache

CH

cluster headache

CI

confidence interval

ICON

Occipital Nerve Stimulation in Medically Intractable Chronic Cluster Headache (trial)

IQR

interquartile range

MICCH

medically intractable chronic cluster headache

MWAFB

mean weekly attack frequency at baseline

ONS

occipital nerve stimulation

RARFB

relative attack reduction from baseline

SD

standard deviation

SF‐36

Medical Outcomes Study (MOS) 36‐Item Short‐Form Health Survey

INTRODUCTION

In many countries, occipital nerve stimulation (ONS) has become an approved and reimbursed therapy for medically intractable chronic cluster headache (MICCH). Several studies and systematic reviews have documented the efficacy, safety, and tolerability. ¹ , ² , ³ , ⁴ , ⁵ Most patients experience subjective (78% are satisfied or very satisfied) and objective (57% show ≥50% reduction in attack frequency from baseline) improvement. ² , ³ Yet, a significant number of patients do not experience sufficient improvement.

Identification of factors that are associated with treatment success would help physicians improve indication for ONS in MICCH. A long‐term follow‐up study showed an association with a low preoperative Hospital Anxiety and Depression Scale‐depression score and long‐term effect, but only reported last follow‐up that was different for each participant. ¹ Another recent report suggested that early onset of cluster headache (CH), chronic cluster headache (CCH), smoking, and seasonal or circadian fluctuations are such associated factors. ⁶ However, this study was small and not adequately controlled, did not use a formal model, efficacy was poorly defined or dichotomized, and follow‐up was only of short duration. ² , ³

Since no reliable associations have been shown, this study aimed to: (i) reproduce these associations, and (ii) identify other associations using formal regression models and strictly defined outcomes in our double‐blind randomized controlled “Occipital Nerve Stimulation in Medically Intractable Chronic Cluster Headache” (ICON) trial, ² and long‐term follow‐up, ³ of the efficacy of ONS in MICCH at 24 weeks, and 2 and 5 years after initiation. We therefore hypothesized that certain baseline clinical characteristics and early treatment responses would be associated with long‐term efficacy of ONS in patients with MICCH.

METHODS

Source of data

In this prospective cohort study, we used data from the investigator‐initiated, international, multicenter, double‐blind randomized controlled ICON trial ² (October 12, 2010, to December 20, 2020) on the efficacy and safety of ONS in MICCH, as defined by the European Headache Federation. ⁷ Its long‐term follow‐up was a 2–8 year prospective open‐label extension of 88 participants in the Netherlands who completed that trial. ³

This is a secondary analysis of the data first described in the long‐term follow‐up study. ³ The analysis was pre‐planned (according to study protocol) but variable selection occurred post hoc.

Participants

Participants in the ICON trial were implanted after a 12‐week baseline period and evaluated double‐blind at 24 weeks, after which they were all followed for another 24 weeks for optimal open‐label treatment. After completion of the ICON trial, Dutch study participants (N = 88) participated in a prospective long‐term follow‐up for up to 9.5 years, completing two Web‐based questionnaires every 6 months, to assess attack frequency and perceived effect. Quality of life was measured with the Medical Outcomes Study (MOS) 36‐Item Short‐Form Health Survey (SF‐36). ³ Inclusion criteria for the ICON trial are listed in Table 1.

TABLE 1.

Inclusion and exclusion criteria for the ICON trial.

Inclusion criteria

International Classification of Headache Disorders‐II (ICHD‐II) criteria for chronic cluster headache (retrospectively all patients also fulfilled the ICHD‐III criteria). Mean attack frequency at baseline of at least 4 attacks/week. Aged ≥18 years. Medically intractable: having failed adequate prior trials of verapamil and lithium and one of methysergide, topiramate, or gabapentin. a Provided informed consent. Agreeing to refrain from starting new prophylactic cluster headache medication, including steroids, or any other therapy aimed at cluster headache, and agreeing to maintain existing prophylactic cluster headache medication from 4 weeks before entering the baseline period throughout the duration of the double‐blind phase of the study. Availability during follow‐up period. A magnetic resonance imaging (MRI) not older than 4 years prior to enrolment must be available to exclude structural lesions potentially causing cluster headache. An exception can be made by the study coordinator in case of stable cluster headache for over 4 years and no change in symptoms after the last MRI scan was performed. Magnetic resonance angiogram of head and neck are to be performed according to study physician's individual judgment.

Exclusion criteria

Other significant neurological or disabling diseases (including other forms of trigeminal autonomic cephalgias), which in the opinion of the clinician may interfere with the study. Pregnancy. Cardiac pacemaker and other neuromodulatory devices. Psychiatric or cognitive disorders and/or behavioral problems, which in the opinion of the clinician may interfere with the study. Taking cluster headache prophylactic medication for conditions other than cluster headache, which in the opinion of the clinician may interfere with the study. Serious drug habituation and/or overuse of acute headache medication (use on ≥10 days/month) for other headaches than cluster headache (excessive use of triptans for cluster headache attacks is not an exclusion criterion). Inability to complete the (electronic) diary in an accurate manner. Structural intracranial or cervical vascular lesions, which may potentially cause cluster headache. Previous destructive surgery involving the C2 or C3 roots (vertebrae) or deep brain stimulation. Enrollment in other clinical studies, which in the opinion of the study clinician may confound the results of this study. Requiring anticoagulation therapy, or antithrombotic, or thrombocyte aggregation inhibitor for a concomitant condition that cannot be stopped perioperatively. The local perioperative protocol of each individual participating center will be followed.

^a All patients fulfilled in retrospect also the consensus criteria of the European Headache Federation for medically intractable chronic cluster headache.

Outcome and included variables

We created two models in which two response measures were used:

1. Relative difference in attack frequency from baseline at 2 and 5 years after implantation

2. Perceived effect at 2 and 5 years after implantation

The following variables were tested for association: age at onset of CCH, time since onset of CCH, sex, smoking, mean weekly attack frequency at baseline (MWAFB), number of ictal autonomic symptoms at baseline, and relative attack reduction from baseline (RRAFB), at 4 and 24 weeks, and 2 years after implantation.

Sample size

No statistical power calculation was conducted prior to the study and the sample size was based on the available data. Since MICCH is relatively rare, all patients who consented to participate in the prospective follow‐up study to the ICON trial were included in the analyses.

Missing data

In case of missing data due to device removal for lack of effect, the last observation was carried forward and the perceived effect was scored as ‘no effect.’ The relative increase in attack frequency was limited to 100% to prevent outliers (two cases after 2 years and one after 5 years).

Ethics

Written informed consent was obtained from all participants and the study protocol of the long‐term follow‐up was approved by the ethical committee of the Leiden University Medical Center (METC‐LDD; Protocol number P10.016).

Statistical analysis

In consultation with a statistician, continuous data were assessed visually (e.g., using histograms). Descriptive statistics are depicted appropriately, i.e., means with standard deviations (SDs), medians with interquartile ranges (IQRs), and counts with percentages, for normally distributed continuous variables, skewed data, and categorical variables, respectively.

A linear regression analysis was performed with RRAFB at 2 and 5 years after implantation as dependent variables, and age at onset of CCH and smoking as independent variables.

Other linear regression analyses were performed with RRAFB at 2 and 5 years after implantation, as dependent variables, and age at onset of CCH, time since onset of CCH, smoking, sex, MWAFB, number of ictal autonomic symptoms at baseline, and the RRAFB at 4 and 24 weeks after implantation (and for the model with RRAFB at 5 years, the RRAFB at 2 years), as independent variables.

The same linear regression analysis was performed with the percentage increase in SF‐36 scores from baseline at 2 and 5 years as a dependent variable. If a regression model showed that it was significantly associated with the dependent variable, the effect of its individual independent variables was calculated.

A binary logistic regression model was used to associate the perceived effect (response or no response, as stated by the participant) of ONS at 2 and 5 years after implantation. Sex and smoking were used as categorical variables (and for the model with perceived effect at 5 years, the perceived effect at 2 years), and age at onset of CCH, time since onset of CCH, MWAFB, the RRAFB at 4 and 24 weeks after implantation, and number of ictal autonomic symptoms at baseline were used as continuous variables.

Finally, a bivariate correlation between age at onset of CCH and time to ONS implantation was analyzed. In all models, all variables were entered simultaneously.

In consultation with a statistician, the assumptions of the linear and binary regression models, including the assumption of linearity, homoscedasticity, the multivariate normality, the absence of multicollinearity, and linearity of logits, were assessed using diagnostics plots (e.g., Q–Q plot and scatter plots), a correlation matrix generated from the model, and a Box‐Tidwell test, when appropriate.

Statistical testing was two‐tailed, considered significant at a p < 0.05, and was not corrected for multiple testing.

Analyses were carried out using the Statistical Package for the Social Sciences (SPSS), version 29.0.2.0. (IBM Corp., Armonk, NY, USA).

RESULTS

Participants

Table 2 shows the baseline data from all 88 participants (66% male (n = 58), mean [SD] age 39 [14] years). Data from 61 (79%) participants were used for the 2‐year analysis and from 42 (48%) participants for the 5‐year analysis. A detailed description regarding follow‐up, loss to follow‐up, and drop‐out rate has been published in the previous reports and can be found in Figure 1. ² , ³

TABLE 2.

Baseline characteristics.

Characteristic	Total	2‐year cohort	5‐year cohort
Participants, n	88	61	42
Sex, male, n (%)	58 (66)	41 (67)	27 (64)
Smoking, n (%)	65 (74)	47 (77)	32 (76)
Autonomic symptoms, n, median (IQR)	5 (3–6)	5 (4–7)	5 (4–6)
Age at CCH debut, years, mean (SD) a	39 (14)	39 (13)	39 (13)
Time since onset of CCH, years, mean (SD) a	6 (5)	6 (5)	5 (4)
Weekly attack frequency at baseline, mean (SD)	19.7 (13.9	20.1 (13.3)	21.2 (15)

Abbreviation: CCH, chronic cluster headache; IQR, interquartile range; SD, standard deviation.

^a The question was asked in whole years.

FIGURE 1.

Flowchart depicting number of active participants in the long‐term follow‐up trial. At the 2‐ and 5‐year timepoints, data from 61 and 42 participants were available (modified from Brandt et al. ³ ).

Difference in attack frequency from baseline

The model in which smoking and age at onset of CCH were used as independent variables was not significantly associated with relative difference in attack frequency after 2 years (R ² = 0.029, adjusted R ² = −0.004; p = 0.417), or 5 years (R ² = 0.034, adjusted R ² = −0.009’ p = 0.458) (Table 3).

TABLE 3.

Linear regression model with smoking and age at onset of chronic cluster headache as independent variables for attack reduction (A) 2 years, and (B) 5 years after implantation.

(A)
	B	95% CI lower	95% CI upper	t	p
Age at onset of CCH	−0.01	−0.02	0.01	−1.14	0.258
Smoking	0.04	0.08	0.17	0.70	0.492

(B)
	B	95% CI lower	95% CI upper	t	p
Age at onset of CCH	0.01	−0.01	0.02	1.19	0.241
Smoking	−0.04	−0.18	0.11	−0.50	0.622

Abbreviation: CCH, chronic cluster headache; CI, confidence interval.

The model in which smoking, age at onset of CCH, time since onset of CCH, sex, MWAFB, number of ictal autonomic symptoms, and relative attack reduction at 4 and 24 weeks, were used as independent variables, was significantly associated with a decrease in relative attack frequency at 2 years (R ² = 0.347, adjusted R ² = 0.247; p = 0.003). Further analysis showed that a 1% increase in relative attack reduction at 24 weeks was significantly associated with 0.45 percentage point increase in relative attack reduction at 2 years (B = 0.45, 95% confidence interval [CI] 0.13–0.76; p = 0.007), and a 1‐year increase in time since onset of CCH was associated with a 4.04 percentage point increase in RRAFB at 2 years (B = 4.04, 95% CI 1.16–6.92; p = 0.007) (Table 4).

TABLE 4.

Linear regression model with smoking, age at onset of chronic cluster headache (CCH), time since onset of CCH, sex, mean weekly attack frequency at baseline, number of ictal autonomic symptoms, and relative attack reduction from baseline at 4 and 24 weeks as independent variables for percentage attack reduction 2 years after implantation.

Variable	Multivariate B (95% CI)	t	p	Univariate B (95% CI)
Age at onset of CCH	−0.43 (−1.60 to 0.73)	−0.74	0.461	−0.71 (−1.90 to 0.49)
Time since onset of CCH	4.04 (1.16 to 6.92)	2.81	0.007*	4.99 (2.48 to 7.50)
Sex (female)	−15.30 (−47.80 to 17.19)	−0.95	0.349	−1.30 (−32.23 to 29.67)
MWAFB	0.12 (−0.97 to 1.22)	0.23	0.821	−0.14 (−1.26 to 0.98)
Number of autonomic symptoms	−1.42 (−8.18 to 5.34)	−0.42	0.676	−1.97 (−8.34 to 4.40)
RARFB at 4 weeks	0.06 (−0.10 to 0.23)	0.74	0.460	0.11 (−0.05 to 0.27)
RARFB at 24 weeks	0.45 (0.13 to 0.76)	2.83	0.007*	0.45 (0.20 to 0.70)
Smoking	−15.23 (−51.48 to 21.02)	−0.84	0.403	−17.65 (−50.97 to 15.67)

Note: The reference category for sex was female. The multivariate B and the univariate B are reported with their 95% CI (lower limit to upper limit).

Abbreviations: CCH, chronic cluster headache; CI, confidence interval; MWAFB, mean weekly attack frequency at baseline; RARFB, relative attack reduction from baseline.

^* p < 0.05

The model in which smoking, age at onset of CCH, time since onset of CCH, sex, MWAFB, number of ictal autonomic symptoms, and relative attack reduction at 4 and 24 weeks and at 2 years were used as independent variables was significantly associated with a decrease in relative attack frequency at 5 years (R ² = 0.546, adjusted R ² = 0.418; p = 0.001). In this model, only a 1% attack reduction at 2 years was associated with a 0.5 percent point attack reduction at 5 years (B = 0.50, 95% CI 0.19–0.82, p = 0.003) (Table 5).

TABLE 5.

Linear regression model with smoking, age at onset of chronic cluster headache (CCH), time since onset of CCH, sex, mean weekly attack frequency at baseline, number of ictal autonomic symptoms, relative attack reduction at 4 weeks, 24 weeks, and 2 years, were used as independent variables for percentage attack reduction 5 years after implantation.

	Multivariate B (95% CI)	t	p	Univariate B (95% CI)
Age at onset of CCH	0.91 (−0.40 to 2.18)	1.46	0.154	0.85 (−0.40 to 2.11)
Time since onset of CCH	0.04 (−3.78 to 3.86)	0.02	0.983	3.47 (0.173 to 6.76)
Sex (female)	−11.56 (−47.05 to 23.93)	−0.66	0.512	−3.89 (−37.59 to 28.82)
MWAFB	−0.30 (−1.38 to 0.77)	−0.57	0.50	−0.85 (−1.89 to 0.19)
Number of autonomic symptoms	2.99 (−426 to 10.23)	0.84	0.408	−0.71 (−7.49 to 6.06)
RARFB at 4 weeks	0.02 (−0.13 to 0.17)	0.24	0.812	0.07 (−0.10 to 0.23)
RARFB at 24 weeks	0.33 (−0.06 to 0.72)	1.74	0.091	0.70 (0.39 to 1.00)
Smoking	−17.55 (−57.84 to 22.75)	−0.89	0.382	−21.43 (−57.82 to 14.96)
RARFB at 2 years	0.50 (0.19 to 0.82)	3.24	0.003*	0.64 (0.43 to 0.84)

Note: The reference category for sex was female. The multivariate B and the univariate B are reported with their 95% CI (lower limit to upper limit).

Abbreviations: CCH, chronic cluster headache; CI, confidence interval; MWAFB, mean weekly attack frequency at baseline; RARFB, relative attack reduction from baseline.

^* p < 0.05

No significant associations for the SF‐36 scores (mental health sum score, physical health sum score, and general health score) were observed, and no correlation between age at onset of CCH and time to ONS implantation was observed.

Perceived effect

When controlling for sex, smoking, age at onset of CCH, time since onset of CCH, MWAFB, the RRAFB at 4 and 24 weeks after implantation (and for the model with perceived affect at 5 years, the perceived effect at 2 years), and number of ictal autonomic symptoms, the adjusted odds for perceiving a positive effect of ONS at 2 years increased by 6% for each additional year of age at CCH onset (adjusted odds ratio 1.06, 95% CI 1.01–1.12; p = 0.033), and 2% for each percentage point attack reduction at 24 weeks (adjusted odds ratio 1.02, 95% CI 1.00–1.04; p = 0.0017) (Table 6).

TABLE 6.

Binary logistic regression model to model the perceived effect (response or no response) of ONS at 2 years after implantation.

Variable	Multivariate aOR (95% CI)	p	Univariate OR (95% CI)
Age at onset of CCH	1.06 (1.01–1.12)	0.033*	1.05 (1.01–1.10)
Time since onset of CCH	1.01 (0.86–1.19)	0.872	1.07 (0.94–1.21)
Sex (female)	0.24 (0.05–1.09)	0.064	0.46 (0.16–1.31)
MWAFB	1.01 (0.96–1.06)	0.724	1.00 (0.96–1.03)
Number of autonomic symptoms	1.17 (0.84–1.62)	0.354	0.99 (0.79–1.24)
RARFB at 4 weeks	0.81 (0.17–3.88)	0.789	1.00 (0.99–1.01)
RARFB at 24 weeks	1.02 (1.00–1.04)	0.017*	1.02 (1.01–1.03)
Smoking	0.28 (0.05–1.63)	0.158	0.47 (0.12–1.81)

Note: Sex and smoking were used as categorical predictors; age at onset of CCH, time since onset of CCH, MWAFB, the RRAFB at 4 and 24 weeks after, and number of ictal autonomic symptoms at baseline were used as continuous predictors. The aOR and its 95% CI (lower limit–upper limit) are reported. The reference category for sex was female. Furthermore, the univariate OR an its 95% CI (lower limit–upper limit) are reported.

Abbreviations: CCH, chronic cluster headache; CI, confidence interval; MWAFB, mean weekly attack frequency at baseline; (a)OR, (adjusted) odds ratio; RARFB, relative attack reduction from baseline.

^* p < 0.05

No factors were observed for perceiving a positive effect at 5 years.

DISCUSSION

Using data from the double‐blind randomized controlled ICON trial, and prospective long‐term follow‐up, we were unable to reproduce the association between smoking and age of onset of CCH, with objective and subjective long‐term effectiveness of ONS for MICCH. In contrast, response at 24 weeks and 2 years appeared to be strongly associated with long‐term response to ONS. Furthermore, a longer duration of CCH prior to ONS treatment was moderately associated with a positive response at 2 years, but not at 5 years. Finally, a later age of onset of CCH was associated with greater treatment satisfaction at 2 years, but not at 5 years. This is consistent with results from another study, ⁶ and from the ICON follow‐up study, showing that nearly three‐quarters of participants who experienced an attack reduction of ≥50% after the first year maintained this response for most of the subsequent 4.2 ± 2.2 years. ³

Possible reasons for this discrepancy are that the study that suggested early onset of CH and CCH, smoking, and seasonal or circadian fluctuations as predictors of efficacy ⁶ was small (26 treated with ONS vs. 88 in this study) and not adequately controlled, did not use a formal model, efficacy was poorly defined (and dichotomized), and the follow‐up was only of short duration. Furthermore, patient selection between the two studies differs: real‐world data vs. stringent randomized controlled trial inclusion criteria. ⁸ As only 19 out of 150 patients were excluded in the original ICON trial, of which nine were excluded due to a low attack frequency, this probably does not account for the observed difference. Additionally, since both ONS protocols used tonic stimulation, stimulation type is unlikely to explain the observed discrepancies. We could not analyze the effect of seasonal or circadian fluctuations because we do not have such data.

The most robust association with long‐term efficacy was early response at 24 weeks and response at 2 years, and 2% increased odds for perceiving a positive effect after 2 years for every percentage attack reduction that was observed at 24 weeks. This raises the question of whether ONS therapy should be continued if patients do not show any attack reduction after 24 weeks. However, we previously showed that 38% of patients who were not a ≥50% responder at the end of the ICON trial (1 year after implantation), still became a ≥50% responder for at least half of the follow‐up period. ³ Furthermore, a discrepancy between attack frequency reduction and subjective response was observed, urging caution when evaluating therapy effect by attack frequency reduction alone. Moreover, despite the evident significant association between attack reduction and the odds of perceiving a positive effect, the magnitude of the effect is moderate. We advise to discuss therapy continuation with the patient and continue the treatment when patients perceive a positive effect, regardless of the effect on attack frequency. Even though the response to ONS therapy may not be optimal, other treatment options are more invasive (deep brain stimulation), but might be considered or have limited evidence for efficacy in MICCH (calcitonin gene‐related peptide monoclonal antibodies, botulinum toxin treatment). ⁵

We should also address some limitations of this study. First, developing a robust, evidence‐based prediction model usually requires a large data set that allows the use of a training and a validation cohort. However, because MICCH and ONS implantation are relatively rare, this is not possible, and we report on association. We report on a relatively large number of patients with MICCH compared with other studies, but the absolute number remains small, warranting caution with its interpretation, with consequently wide CIs and an increased risk of overfitting. Second, consistency in the interventional procedure is also critical, but despite efforts to standardize ONS surgery, differences in lead placement persist. Furthermore, anatomical differences in the location of the greater occipital nerve may result in different distances of leads to the greater occipital nerve between patients, potentially influencing the intended effect, and further complicating prediction. ⁹ Therefore, most studies to date have failed to identify predictors of, or factors associated with, treatment success. ¹ , ¹⁰

Third, we need to address the issue of missing data, as we have previously described in the long‐term follow‐up study ³ : “Due to the staggered inclusion, missing data is unavoidable. The reasons for missingness should be considered very carefully as non‐random missingness could bias the results. Most of the missing data after 5 years of follow‐up is due to the fact that, because of the staggered inclusion, these participants had not yet reached 5 years of follow‐up after completion of the ICON trial when the study ended (n = 24 cases). These data can therefore be regarded as ‘missing completely at random.’” In the event of study termination due to lack of effect, or attack freedom, it is reasonable to assume that the missingness can be explained by reasons about which we have full information (i.e., attack frequency and quality of life). We therefore assume that the reason for missingness is likely present in the data points before the missing data (lack of effect or complete remission). Accordingly, we feel that the missingness should be labeled as “missing at random rather than not at random.” ³ Additionally, since we wanted to account for lack of effect in the results, we decided to use the last observation carried forward for the participants who had missing data because of device removal due to lack of effect.

Fourth, with only three measurement points (baseline, 2 years, and 5 years), the described data may not capture non‐linear changes that occur between assessments, potentially oversimplifying the trajectory of each participant's response. As previously described, roughly one third of participants showed a shift in responder status over the course of the long‐term follow‐up period. ³ Finally, although this study began as a randomized controlled trial, the open‐label follow‐up phase means that the later data are observational in nature, so any uncontrolled confounding during this period could influence the findings.

In conclusion, in a controlled setting, early onset of CH, CCH, and smoking were not associated with the treatment of ONS for MICCH as previously suggested by others in an uncontrolled setting. Early response at 24 weeks after the initiation of ONS was the only factor that was associated with long‐term efficacy, which was identified. Since previous studies have shown that a large proportion of patients with MICCH improve with ONS and we did not show any reliable patient characteristics that were associated with treatment efficacy, we recommend offering ONS to all patients with MICCH.

AUTHOR CONTRIBUTIONS

Roemer B. Brandt: Conceptualization; data curation; formal analysis; methodology; validation; visualization; writing – original draft; writing – review and editing. Casper S. Lansbergen: Writing – original draft; writing – review and editing. Linda Kollenburg: Writing – review and editing. Cecile C. de Vos: Methodology; supervision; writing – original draft; writing – review and editing. Michel D. Ferrari: Conceptualization; methodology; supervision; writing – original draft; writing – review and editing. Frank J. P. M. Huygen: Conceptualization; supervision; writing – original draft; writing – review and editing. Rolf Fronczek: Conceptualization; methodology; supervision; writing – original draft; writing – review and editing.

CONFLICT OF INTEREST STATEMENT

Roemer B. Brandt, Casper S. Lansbergen, Linda Kollenburg, Michel D. Ferrari, Frank J. P. M. Huygen, Cecile C. de Vos, and Rolf Fronczek report no relevant conflicts of interest.

APPENDIX 1.

The ICON trial study group

Investigators are listed by center

1. Leiden University Medical Center: M.D. Ferrari (Chair), L.A. Wilbrink, I. F. de Coo, P.G.G. Doesborg, E.C. Bartels, E.W. van Zwet

2. Erasmus Medical Center: F.J.P.M. Huygen (Vice Chair)

3. Canisius Wilhelmina Hospital: W. Mulleners, E. Kurt

4. Radboud Medical Center: R.T.M. van Dongen

5. Zuyderland Hospital: O.P. Teernstra, P.J.J. Koehler, G.H. Spincemaille, L.A. Wilbrink

6. Diakonessenhuis Zeist: F. Wille

7. Alrijne Hospital: K. Burger, J. Haan

8. Boerhaave Medical Center: E.G.M. Couturier

9. Rijnstate Hospital Arnhem: J.W. Kallewaard

10. University of Twente: Peter H. Veltink

11. Medtronic BV: R. Buschman

Brandt RB, Lansbergen CS, Kollenburg L, et al. Factors associated with efficacy of occipital nerve stimulation in medically intractable chronic cluster headache. Headache. 2025;65:1626‐1633. doi: 10.1111/head.14985