We evaluated approaches to the AT of migraine by physicians, identifying clinical characteristics, treatment patterns, and patient-reported outcomes in patients who were responders and insufficient responders to AT in China. We found that, compared with responders, insufficient responders had different clinical characteristics, and carried a greater disease burden, yet were treated similarly by their physicians.

In this study, patient demographics were generally representative of the migraine population in China [5, 27]. Insufficient responders to AT comprised 44.0% of the study population, which is higher than the patient proportions (30–40%) reported in studies using a similar definition of insufficient response [11, 28]. Insufficient responders mainly had symptoms of moderate severity, and more commonly experienced migraine with aura, MOH, and unilateral pain than responders; responders had more mild symptoms, and more commonly experienced migraine without aura and bilateral pain. These characteristics differ from those in US patients with migraine, where migraine with aura is more common, particularly in responders, MOH is over twice as common, and bilateral pain is more often experienced by insufficient responders [11].

It is generally known that the migraine-related disability increases with headache frequency [29, 30]. The present data indicated that for responders and insufficient responders both the average number of headache days per month was very low (below 4) and that the reporting of disability by patients in terms of MIDAS grading was also low. The DSP methodology whereby physicians completed patient records forms for their next prospectively consulting patients, as well as the absence of quotas on patient type, means that the patient cohort is representative of the consulting population. Therefore, the majority of patients reporting little/no or mild disability in the MIDAS is in line with what would be expected from patients with a low frequency of migraine.

While the use of analgesics is justifiable for migraine, it is widely known that headache chronification occurs with their overuse [31]. In our study, MOH was significantly more likely to be reported by insufficient responders than responders. This may be due to treatment inefficacy, whereby insufficient responders do not receive pain relief on their current medication and therefore take their medication more frequently than recommended. Such medication overuse causes changes in pain processing networks, dependency networks and sensitization in patients with an underlying susceptibility for migraine progression [31].

Notably, more responders lived with another family member, which was found to be a significant indicator that a patient would respond to treatment. Better treatment effects could result from patients living with another family member who could provide social support and care. Indeed, patients with pain with supportive families reported significantly less pain intensity, greater activity levels, and tended to be working compared with patients with a non-supportive family [32].

This analysis found that migraine treatment patterns were overall similar between responders and insufficient responders, with NSAIDs, (mostly ibuprofen and diclofenac) most frequently prescribed currently. This similarity in treatment pattern may be because insufficient responders were not receiving an AT that could adequately manage their migraine attacks, indicating their unmet need. AT prescribing has also been reported to be similar between insufficient and sufficient responders to AT for migraine in Japan [12]. Current use of triptans did not significantly differ between the groups. Also reflecting their unmet treatment need, insufficient responders had poorer adherence to treatment during a migraine attack than responders.

Despite some consistency of the recommended treatments with American and European guidelines, treatment patterns in China generally differed from those in the US [11]. Data retrieved from the CHIRA outpatient database in 2016/2017 found that 26.4% of patients were prescribed acute medication. Similar to our findings, 68.8% of patients received non-aspirin NSAIDs (mostly ibuprofen), 7.1% opioids, 6.1% ergotamines, and 3.3% of patients received triptans [5]. The common prescribing of NSAIDs as acute medications may reflect their accessibility and relatively low cost. The notable low triptan use among Chinese patients contrasts with the high use in the US (> 80%) [11] and in Japan (> 75%) [12, 33]. The low rate of triptan use in China may be due to limited triptan availability in many hospitals and pharmacies resulting in barriers to access [5]. It may also be explained by their high price limiting their widespread use [5, 34]. Furthermore, compared with the seven types of triptans and their different forms in the US, the choice of triptans is small in China, with only three triptan types commercially available (sumatriptan, zolmitriptan, and rizatriptan) [5]. Triptan under-use may be through patient choice, with most patients unable to afford the high cost of each triptan tablet.

In contrast with the US [11, 35], the low opioid use in our study reflects the overall low opioid use in China due to government regulations [5]. While effective pain relief and guidelines may predominate in the physicians’ treatment decision, factors associated with patient choice and medication use, such as cost, availability of medicine and tradition (i.e., a culture’s customs, ideas and beliefs), are also important because these factors could influence the treatment decision.

Overall, the majority of patients in the US take their medicine at the first sign of a migraine attack [11] while patients in China wait until or even after the pain starts. Such behavior may suggest culture differences. A US study found that patients taking medication for a migraine attack when the pain starts were significantly (p < 0.001) more likely to be insufficient responders than those taking their medication at the first sign of a migraine attack (i.e., during the premonitory phase) [11]. This was not found in our Chinese study. The majority of patients in both groups waited for the pain to start before taking their medication. A possible reason for this difference may be the greater use of triptans in the US (82.8%) [11], which provide greater efficacy when taken while migraine pain is still mild [36].

While this analysis showed that a significantly greater proportion of insufficient responders vs. responders did not experience side effects of AT, the impact of side effects on insufficient responders’ lives was greater than that on responders. Generally, differentiating between treatment-related adverse events and migraine symptoms can be difficult, particularly when the event could be either, such as in the cases of nausea, vomiting, and dizziness. Thus, a moderate and severe impact on insufficient responders’ lives could be due to migraine symptoms while the lack of impact on responders’ lives could be due to treatment efficacy. On the other hand, a high response rate generally indicates that the drug is effective and thus might cause side effects. Furthermore, it is reasonable to associate a lower rate of side effects with reduced effectiveness, since some migraine symptoms like vomiting might result in decreased drug absorption, which in turn may contribute to reduced efficacy and fewer side effects.

Several studies have shown that migraine negatively affects QoL [37]. Insufficient responders experienced more severe symptoms, greater disease burden, worse QoL, and greater impairments in work productivity and activities versus responders. Corresponding with our findings, more severe workplace problems has been associated with a higher disability level in patients with migraine [38], and people with migraine generally continue to work despite symptoms, with considerable productivity loss due to presenteeism [39]. In contrast to our analysis of Chinese patients, data from the US Adelphi Migraine DSP found that patients with migraine report worse disability (MIDAS; insufficient responders only), with proportionally more insufficient responders with moderate or severe disability but a lower impact of headache on work (WPAI; responders and insufficient responders) [11]. However, patients in China reported worse disability, with proportionally more with mild or moderate disability, and worse impairment while working, overall productivity, and activity, compared to patients in Japan [12]. These findings may also be due to cultural differences between Asian and Western societies in dealing with illness.

Based on bivariate analyses, among the multiple comparisons, Chinese patients who were insufficient responders achieved pain freedom within 2 h in ≤ 3 of five attacks, experienced migraine with aura, had unilateral pain as a symptom of an attack and as a most troublesome symptom, sensitivity to smell as a symptom of an attack, more moderate symptoms, gained partial symptom control, and reported greater side-effect impact. After consideration of the variables identified by LASSO, these multivariate analyses confirmed that Chinese patients who were insufficient responders were significantly more likely than responders to have diabetes, unilateral pain, vomiting, sensitivity to smell, visual aura/sight disturbance, and worse disability (increase in MIDAS total score).

Despite the association of migraine with several comorbidities and metabolic/endocrine disorders, including insulin-related factors, the relationship between migraine and type 2 diabetes mellitus remains unclear due to relatively few existing studies and conflicting results [40, 41]. Several mechanisms have been suggested for an inverse relationship between migraine and diabetes [41], including lifestyle, environmental, and genetic factors, since migraine can be triggered by such factors; the actions of biochemical biomarkers relevant to migraine pathophysiology (e.g., proinflammatory cytokines and neuropeptides); and the autonomic nervous system since it influences the metabolic changes that occur. However, the role of diabetes in the protection against migraine attacks remains uncertain [41].

The LASSO procedure is a regularization method that supports simple models with fewer parameters. It is useful in cases with limited evidence because it enables the testing of many potentially correlated factors and identifies those most important. The present outcomes reflect the relative strength of the factors in each model, but they do not suggest clinical meaningfulness, and they ignore nonsignificant variables that may otherwise be important.

The AMPP study identified a different set of significant predictors of inadequate treatment response to those in our analysis, [42], and subsequently found yet a different set of predictors [11]. A high MIDAS total score (severe disability) was the only predictor found to be similar among those identified in our analysis of Chinese patients. Factors identified in our analysis also differ from those found in the Japan Adelphi Migraine DSP population [12]. Such differences between real-world studies are likely due to varying patient characteristics, behaviors, and individual treatments, and time of data capture.

While outside the scope of this manuscript, future research in the differences in response patterns between patients with aura and those without in China and elsewhere would be valuable.

The point-in-time design of the DSP prevents any conclusions about causal relationships, although identification of statistically significant associations is possible. It therefore cannot be determined if the patient characteristics caused the insufficient response to AT or vice versa. Physicians were selected based on the number of patients with migraine seen; therefore, the physicians were experienced with treating migraine and their patient load and standard of care may not reflect those of more general practitioners. Generalization of the findings to all patients with migraine may also be limited as the analysis may represent the patient burden and outcomes of patients who visit their physicians and who may be more severely affected by migraine than patients who do not consult their physician as frequently. Despite such limitations, this collection of real-world data enabled evaluation of typical variations in AT prescribing patterns, migraine-related symptoms and burden, and satisfaction with AT.

A strength of this study is that it uses real-world data collected from the Adelphi Migraine DSP that uses standardized methodology and is well validated. Nevertheless, our findings should be considered with regard to the limitations of the DSP survey. While the data were collected several years ago (2014), they are still relevant considering that the treatment paradigm has not changed and no new migraine treatments have since been approved in China. However, data were collected over a relatively short time period (6 months), which may have influenced the findings.