Long-term outcomes of occipital nerve stimulation for chronic migraine: a cohort of 53 patients

Weiner and Reed were the first to report the potential use of ONS for intractable occipital neuralgia [36]. Subsequent review and imaging of the patients by headache specialists, however, suggested many of these patients actually had CM [22]. There have now been three placebo-controlled studies on ONS in CM and although their results have been somewhat contradictory recent meta-analysis suggests an overall positive effect of treatment [17, 30, 34, 37]. The PRISM (Precision Implantable Stimulator for Migraine) study conducted by Lipton et al., only available in abstract form, failed to show a significant difference in the reduction of migraine days/month between the active and sham groups after 12 weeks (−5.5 vs. 3.9 days/month, p = 0.29) [17]. Saper et al. published results of the ONSTIM (Occipital nerve stimulation for the treatment of intractable chronic migraine headache) trial in 2010 [30]. Of the 77 patients included, all subjects had failed to respond to at least two different classes of medication and had reported positive response to greater occipital nerve block. Various outcome measures including reduction in headache days, pain intensity and pain duration were numerically superior in the treatment group. A responder was defined in this study as a subject reporting a more than 50 % reduction in monthly headache days or a more than three-point reduction in average pain intensity. Three-month responder rates were 39 % for the active group, 6 % for the sham-control group and 0 % for the medical management group. The percentage reduction in severe headache days a month was 24.4 % (±43.6) for the active group and 10.3 % (±34.0) for the sham group, corresponding to a reduction of 5.1 days (±8.7) a month in the active group and 2.2 (±6.4) in the sham group. The largest randomized sham-controlled study on ONS in CM was conducted on 157 subjects by Silberstein et al. [34]. This group failed to find a significant difference in the primary outcome measure (those achieving a more than 50 % reduction in average pain intensity at 12 weeks) between active (17.1 %) and sham (13.1 %) stimulation groups. However, a number of secondary outcomes did suggest that ONS had a benefit including the numbers achieving a 30 % reduction in pain severity and a 30 % reduction in headache frequency. Long-term open-label follow-up of this cohort for a total of 12 months revealed that there was a significant reduction in the number of headache days (defined as days with more than 4 h of moderate-to-severe pain) of 6.7 days (±8.4) [10]. The percentage reporting a more than 50 % reduction in headache days and/or pain intensity was 47.8 %. Pooled results from these three trials show that ONS is associated with a mean reduction of 2.59 moderate-to-severe headache days a month (95 % CI 0.91, 4.27) after 3 months treatment compared with sham control [8].

Several open-label series have been published suggesting efficacy of ONS in CM, however, many of these have been of small numbers with restricted follow-up. Two centers analyzed their long-term data in retrospective reviews. Brewer et al. reported 12 CM patients with average follow-up of 34 months (range 1–70 months) and Palmisani et al. reported 19 CM patients with average follow-up of 10 years (range 1–19 years) [5, 29]. Outcome in both was based on the patients subjective reporting via a telephone interview. In the Brewer series, five patients (41.6 %) reported a more than 50 % overall benefit whereas nine (47.3 %) of the Palmisani series reported a more than 50 % reduction in pain intensity and/or frequency.

Our uncontrolled, open-label, prospective observational study showed that in highly refractory CM patients with prolonged follow-up, ONS resulted in a significant reduction of moderate-to-severe headache days a month (8.51 days). In total, 45.3 % of patients showed a more than 30 % reduction of monthly moderate-to-severe headache days following treatment. Five patients were completely pain free and had been for prolonged periods. Significant improvements were also seen in pain intensity, daily pain duration and headache related disability. Although our response rate is below that of the open label series quoted above, our outcome measure is objective and from prospectively completed headache diaries and is thus a more robust measure. The use of a 30 % improvement in outcome measure is accepted in the chronic pain literature as representing a “much improved” clinical state and has been accepted by both the International Headache Society clinical trials subcommittee as being a realistic and clinically relevant improvement in those with chronic migraine [10, 11, 33]. With this is mind and with such a complex group of patients present in this cohort, we feel that a 30 % improvement level is justified.

The primary outcome measure of moderate-to-severe headache days was chosen in accordance with the 2008 guidelines for controlled trials of prophylactic treatment of chronic migraine in adults produced by the Clinical Trials Subcommittee of the International Headache Society [33]. However, a wide variety of headache outcome measures (e.g. “headache days”, pain intensity) have been used in previous case series of ONS for CM and so direct comparison between studies can be somewhat difficult. With this in mind, there is an obvious need for a consensus on the most appropriate outcome measures for ONS efficacy. Our group is also more complicated than those in the previous series with many suffering multiple chronic headache conditions, reporting a higher number of failed past medications and with 94.3 % of them recording background or interictal pain. In such an intractable group of patients, a continued response in over one-third of patients after such prolonged follow-up should be viewed with cautious optimism.

The rates of serious adverse event in our series was below that of previous reported groups. In the randomized trials, concerns were raised over the high rates of lead migration and infection. The ONSTIM trial quoted rates of lead migration at 24 % and infection in at least 18 % of patients, whilst the Silberstein study reported rates of 19 and 7 %, respectively. Our group had no episodes of lead migration and only a single episode of mild wound site infection treated with oral antibiotics. Our implants were conducted by a single highly skilled surgical team and our results mirror those found by Sharan et al. in describing high levels of implanter experience being associated with significantly lower levels of complications [32]. Recent guidelines recommending that ONS should only be carried out in a limited number of highly specialized centers should lead to improvements in adverse event rates and a reduction in the current discrepancy between centres.

Our group do not conduct trial-periods of stimulation as is carried out in a number of other centres. The intention of such trials is that they will positively select those most likely to respond to long-term ONS treatment and they have in-fact been used as inclusion criteria for a number of controlled trials. However, it is clear that a positive trial does not guarantee longer-term success. In the study by Silberstein et al., despite all subjects having a positive trial period, the study still failed to reach its primary endpoint [34]. Other open-label series support the view that trial stimulation does not predict success. Palmisani et al. had a trial success rate of 88 % but removed 7/23 systems implanted due to lack of efficacy and Brewer et al. reported high trial success rates (89 %) but a long-term benefit in only 42 % CM patients [5, 29]. A study using longer trial-periods of one month has also failed to show an association between trial response and ONS outcome [28]. Our data suggests that there is a delay in patients reporting clinical effect of ONS which may be up to 6 months in responders. This may explain why trial stimulation does not predict success and suggests that the early response reported by some groups may be due to a placebo effect. The hypothesis that the neuromodulatory effects of ONS are due to slow, plastic changes within the pain-structures of the brain is supported by our data on time to clinical effect and observation of gradual return of pain when ONS is stopped. This hypothesis of plastic-change is in direct conflict to the rapid action seen in trial stimulation. Given that response rates in those groups using trial stimulation are similar to our patients who do not undergo such trials, we do not feel that the current evidence supports the use of trial stimulation and the additional surgical risks this entails. This remains a controversial area and we support further investigation into the true predictive value of trial stimulation in ONS.

Assessments of headache related disability and quality of life showed numerical improvements but only HIT-6 and EQ-VAS showed any statistical improvement in the cohort as a whole. In a subgroup analysis, however, those with a positive response were found to have statistically significant reductions in a variety of quality of life and affect measures which were not mirrored in the non-responder group. A failure to observe significant change across all assessments despite improvements in headache frequency is reported in cases of epilepsy surgery and spinal cord stimulation and has been attributed to a “burden of normality” [1, 20]. Given that 33.9 % also suffered other headache types that did not necessarily respond completely to ONS and that nearly all patients continued to have migraine pain of some level, they will still exhibit a disability burden from their pain, even if their migraine has significantly improved. This theory is supported by a lower reduction in disability scores in the multiple phenotype group compared to the CM alone group of patients. Recently, Clark et al. reported on the long term functional outcomes of combined supra-orbital and occipital nerve stimulators for CM [9]. The group found that improvements in functional outcome (MIDAS and BDI) were only significant during the first 6-months post implant but not after prolonged follow-up (average 44.5 months). They speculated that this was due to the loss of a “honeymoon period” and as yet unexplained complex interactions between pain and functional status. Our cohort seems to suggest that this is not necessarily accurate as even after a follow-up period of nearly 4 years, over a third of patients still reported clinical response and significantly reduced headache disability.

The strengths of this study include the large sample size, the long follow up period and, importantly, the prospective nature of the data collection (a first in long-term observational ONS cohorts of CM). The real-life nature of the data is also valuable. Patients were not subjected to the strict inclusion criteria of a study and represent the types of highly complex CM patient typically seen in specialized neuromodulation centers. The limitations of this study are mainly centered on the lack of a placebo or sham-stimulation group. However, it is most unlikely that our observations can be explained by placebo alone. We found that there was a delay of months to reporting clinical effect (5.50 months) and a delay before pain worsened when the device was off (2.50 months). These observations are reproducible across multiple ONS cohorts for a variety of primary headache disorders and argue against a pure placebo effect [6, 16, 19]. This time delay may also explain why shorter trials, reporting at 3-months post-implant, do not mirror the more favorable open-label clinical experience of ONS in CM. Other factors against a pure placebo response include the previous intractable nature of the group, a stable response with long-term follow up and the previously quoted placebo rates of between 6 and 13 % in the controlled trials of ONS in CM being below the 45.3 % response rate we quote here [30, 34]. The extrapolation of 2-weeks diary data to represent a month is not ideal but the time-span was chosen as it is the normal diary kept by all of our patients seen in clinic, with or without ONS. The data was collected from patients in a real clinical environment and we are aware that asking for too much information may lead to patients being unable to comply with requirements. Therefore, a 2-week diary was chosen to ensure high rates of compliance and diary completion. This method may be more at risk of being influenced by natural fluctuations in CM severity, however, in our cohort of highly refractory patients such fluctuations were not commonly seen prior to ONS. In the future, consensus on the most relevant outcome-measures for ONS and the development of electronic-diaries may improve ease of dairy keeping and allow long periods of data to be collected easily.

The high levels of complications requiring surgical intervention published in the literature have led to concerns over the cost-effectiveness as well as safety of the procedure. The current equipment used for ONS is designed for spinal cord stimulation and not intended for implantation in the occipital region. Advances in technology have already led to reductions in intervention rates, for example a reduction in need for battery replacement with rechargeable IPG development, and hopefully ONS specific equipment may be available in the future. However, it must be noted that the one company (St Jude Medical) who were granted a European CE Mark Approval for the use of their ONS to treat CM in 2012 had that approval removed in 2014 as it was felt that there was not enough data to demonstrate that the benefits outweighed the risk of therapy [35]. These concerns are a major issue in neuromodulation for migraine and raise the need for high quality, well planned, large placebo-controlled trials to look at efficacy and safety in the long-term treatment of CM.